SAVEGRP(8)SAVEGRP(8)NAMEsavegrp - start a group of NetWorker clients saving their filesystems
SYNOPSISsavegrp [ options ] [ -R | -G ] [ groupname ]
options: [ -EIOFXmnpv ] [ -l level | -C schedule ] [ -N parallelism ] [
-e expiration ] [ -w browse ] [ -y retention ] [ -t date ] [
-r retries ] [ -P printer ] [ -W width ] [ -b backup snapshot
] [ -c client [ -c client ... ] ]
DESCRIPTION
The savegrp command runs a group of NetWorker clients through the
process of saving their filesystems (using save(8)). The group of
clients is selected by naming a NetWorker group (see nsr_group(5)),
from which individual clients can be selected by using one or more -c
options. If no group name is specified, the NetWorker group Default is
used. If a NetWorker group is named, clients whose nsr_client(5)
resources specify the named group in their group attribute will be
saved. If an explicit client list is also specified, savegrp will only
back up those clients, with respect to the named group. The savegrp
command will automatically make a clone of the newly saved data when
the appropriate attributes are set on the NSR group resource (see
below).
The savegrp command is normally run automatically by nsrd(8), as speci‐
fied by each group's nsr_group(5) resource.
The savegrp command will set up an RPC connection to nsrjobd(8) to
request execution of a save(8) job on each client for each filesystem
listed in the nsr_client(5) resource save set attribute. If a save set
of All is specified for a client, savegrp will request from the client
a list of filesystems to be saved (this is called the probe operation).
The probe expands All into a list by looking for filesystems that are
both local and automatically mounted on that client machine (e.g. NFS
mount points and filesystems mounted manually are generally ignored).
The exact determination of which filesystems to save varies between
different operating systems. See savefs(8) for additional details on
the probe operation. To see which filesystems a client saves, run a
savegrp preview, savegrp-c client -p (assuming the client is in the
Default group). Each filesystem saved is called a save set.
For NDMP clients, savegrp will run backup command nsrndmp_save(8) given
in the client resource on the NetWorker server. If the command line
options of nsrndmp_save has -I hostname, then savegrp will run
nsrndmp_save(8) command on the given hostname. For more command line
options and details of nsrndmp_save, see nsrndmp_save(8).
The savegrp command attempts to keep multiple clients busy by queuing
all the save jobs immediately for the client's queue maintained by nsr‐
jobd(8).
The parallelism attribute in the nsr_service(5) resource is the maximum
number of save sets to run simultaneously. Modifications to this
parameter will take effect as save sets complete - if the value is
reduced, no new save set will be started until the number of active
save sets running drops below the new value.
When the savegrp is started from the command line, it does not automat‐
ically pick up the level attribute specified in the group resource.
However, when -l level otion is explicitly specified, the savegrp
command performs the requested level backup.
When all of the save sets are completed on a client for the group, the
client's index on the NetWorker server are saved. If the client is
scheduled for level=incr index backup is done at "-l 9"(level 9) with
"-f -". The default directive causes the backup of all files with
extension "rec" under client's index directory(/nsr/index/<client‐
name>/db6 on a Unix machine) to happen. The manpage for save(8) has
details of "-l" and "-f" options. If the NetWorker server is one of
the machines being saved, NetWorker bootstrap is saved after all the
other clients are completely done. When the server's bootstrap is
saved, the bootstrap save set information is printed to the default
printer (or another specified printer). If savegrp detects that the
NetWorker server is not listed in any active group (a group with its
autostart attribute set), then the server's bootstrap is saved with
every group.
The savegrp command detects other active invocations of the same group,
and exits with an error message. If two different NetWorker groups are
running simultaneously, they each will run up to respective group's
parallelism and not overrun the server and client parallelism ; as the
nsrjobd (8) will control the parallelism and hold these backup jobs in
a queue before starting. NetWorker server also controls the device
parallelism of these sessions to write to the backup devices at a time.
The progress of the actively saving clients can be monitored using the
Java based NetWorker Management Console or the curses(3X) based nsr‐
watch(8) program. The NetWorker Management Console or nsradmin(8)
browser may also be used to examine the completion status and work list
of each NSR group resource, and allow you to track the progress of each
savegrp. These two attributes allow you to track the progress of each
savegrp. See nsr_group(5) for more details.
When savegrp starts, it sends an NSR notification (see nsr_notifica‐
tion(5)) with an event of savegrp and priority of info to the NSR noti‐
fication system. This event is normally logged in the messages
attribute of the nsr_service(5) resource, and in the log file specified
in the Log default NSR notification resource.
The save sets are automatically cloned, when all the save sets have
finished, and the NSR group resource has the clones attribute enabled.
The client save sets and their indexes are cloned before the bootstrap
save set is generated so the bootstrap information can track both the
original set of save sets and their clones. The bootstrap save set is
also cloned. Clones will be sent to the pool named in the clone pool
attribute. Changing the values of these attributes while savegrp is
running has no effect; they must be set before savegrp starts. The
nsrclone(8) command is used to clone the save sets. savegrp uses a
heuristic to determine which save sets were generated as part of the
group; it may occasionally clone more save sets than expected, if a
client has its filesystems separated into multiple groups that run at
the same time. Note that at least two enabled devices are required to
clone save sets and/or atleast two enabled ndmp devices are required to
clone ndmp save sets.
savegrp, when started for snapshot enabled groups, creates snapshots
for each of the clients configured in the group resource (see
nsr_group.5). If any of the save sets, for clients configured in this
group resource, is non-snapshot capable, then savegrp will report a
failure while trying to create the snapshot. If any of the client
resources, configured with this group, has the Keyword All (see
nsr_client.5), then the non-snapshot capable file systems on the client
node will be ignored and no error message is generated for their fail‐
ures.
NOTE: This option is available with EMC's PowerSnap Module only.
When the save sets are all complete and cloned (if cloning is enabled),
an NSR notification with an event of savegrp and priority of notice is
sent to the NSR notification system. This is generally set up to cause
e-mail to be sent to the root user specifying the list of clients who
failed (if any), and all the output collected from all clients. The
format and common error messages included in the savegrp notification
are explained in the SAVEGRP COMPLETION NOTIFICATION MESSAGE section.
OPTIONS-E Causes save(8) on each client to estimate the amount of data
which will be generated by each save set before performing it.
This will result in the filesystem trees being walked twice -
once to generate a estimate of how much data would be generated,
and again to generate a save stream to the NetWorker server.
Note that the data is only read from the disk on the final
filesystem walk, as the estimate is performed by using inode
information.
-I Disables the saving of each client's index.
-O Saves only each client's index (the bootstrap is also saved).
-m Disable monitor status reporting, including all NSR notification
actions. When this option is selected, the progress or comple‐
tion of save operation is not reported. The notification of
bootstrap information is not affected by -m option.
-n No save. Cause save to perform an estimate as described for -E,
but not to perform any actual saves. This option also sets -m.
-p Runs the probe step on each client, so you can see which
filesystem would be saved and at what level, but do not actually
save any data. This option also sets -m. The output generated
by the -p option may show several save levels for each save set
at different points in the output, as savegrp learns the correct
level. This is the expected behavior, and can be useful for
debugging. The actual level the savegrp uses is shown the last
time each save set is displayed in the output. The media pool
the save set would be directed to is also listed in the preview
output.
-v Verbose. Prints extra information about what savegrp is doing.
The -q flag is also not passed to the save command.
-G Run the group; apply no restart semantics. This is the default
mode of operation; the option is provided for compatibility with
other versions of savegrp.
-R Restart. This option is used to restart a group that was
stopped or if savesets failed and they need to be retried. The
restart window attribute of the group is used to determine if it
is too late to be restarted. If the window has elapsed the
restart is converted into fresh start.
-l level
The level of save (see nsr_schedule(5)) to perform on each
client. This overrides the save level which savegrp would nor‐
mally automatically determine. -l and -C cannot be specified
together.
-C schedule
The name of the NSR schedule (see nsr_schedule(5)) to be used in
the automatic save level selection process which savegrp nor‐
mally performs. This overrides the save schedule which savegrp
would normally use for a given client. -l and -C cannot be
specified together.
-e expiration
Set the date (in nsr_getdate(3) format) when the saved data will
expire. When a save set has an explicit expiration date, the
save set remains both browsable and non-recyclable until it
expires. After it expires and it has passed its browse time,
its state will become non-browsable. If it has expired and it
has passed its retention time, the save set will become recy‐
clable. The special value forever is used to indicate that a
volume that never expires (i.e. an archive volume) must be used.
By default, no explicit expiration date is used.
-w browse
Sets the date (in nsr_getdate(3) format) after which the saved
data will no longer be browsable. By default, the server deter‐
mines the browse date for the save set based on the browse poli‐
cies in effect. This option allows overriding the existing
policies on a save by save basis.
-y retention
Sets the date (in nsr_getdate(3) format) when the saved data
will become recyclable. The special value forever is used to
indicate that a volume that never expires (i.e. an archive vol‐
ume) must be used. By default, the server determines this date
for the save set based on the retention policies in effect.
This option allows overriding the existing policies on a save by
save basis.
-t date
The time to use instead of the current time for determining
which level to use for this savegrp (in nsr_getdate(3) format).
By default, the current time is used.
-F Automatically perform a full level backup if save set consolida‐
tion fails. This option is ignored if the backup level is not
"c".
-X Automatically remove the level 1 save set after save set consol‐
idation builds a full level save set. This option is ignored if
the backup level is not "c". It is also ignored if the backup
level is "c" but the save set consolidation process fails.
-r retries
The number of times failed clients should be retried before
savegrp gives up and declares them failed. The default is taken
from the group resource. Abandoned saves are not retried,
because they may eventually complete. Retries are not attempted
if -p is specified.
-P printer
The printer which savegrp should use for printing bootstrap
information.
-W width
The width used when formatting output or notification messages.
By default, this is 80.
group Specifies the NetWorker group of clients that should be started,
rather than the default NSR group (which has the name attribute
of default). See nsr_group(5) for more details.
-b backup snapshot
This option should be used only with snapshot groups. Using this
option with non-snapshot groups causes this option to be
ignored. When passed for a snapshot group, this option will
configure the specified snapshots to be backed up to tertiary
storage.
-c client
The name of a client on which to save filesystems. There can be
multiple -c client specifications. When -c options are speci‐
fied, only the named clients from the specified group (which is
"Default" if no group is specified) will be run.
-N parallelism
The parallelism value overrides any other parallelism considera‐
tions that savegrp may use to avoid over-utilizing the system's
resources.
RESOURCE TYPES
NSR Use the parallelism attribute for the maximum number of
saves to start simultaneously.
NSR group The attribute work list contains values in groups of 3,
specifying the client name, level of save, and path to
save, for each save set not yet completed. The attribute
completion contains values in groups of 4, specifying the
client name, path saved, status, and the output, for each
save set completed. The success threshold attribute con‐
tains the threshold to determine the success of all
savesets within the group.
NSR schedule Used by the savegrp command with each client's
nsr_client(5) resource to determine which level of save to
perform for each specified save set.
NSR client Each client resource names the groups it should be saved
by, the names of the save sets which should be saved, the
name of the schedule to use (see nsr_schedule(5)) and the
name of the directives to use (see nsr_directive(5)).
NSR notification
Three kinds of notices are sent to the NSR notification
system, both with the event attribute of savegrp. While a
savegrp is in progress, status notices are sent with the
priority of info. At completion of a savegrp, a notice is
sent containing the collected output of all saves, and the
name of clients which had a save which failed (if any).
This notice will have an event type of savegrp, and a pri‐
ority of notice. If savegrp is interrupted, a notice
stating the group was terminated, with an event type of
savegrp, and a priority of alert will be sent. These last
two typically will result in the notice being encapsulated
in a mail message to root.
SAVEGROUP COMPLETION NOTIFICATION MESSAGE
The savegroup completion notification message contains 6 parts: the
header, the Never Started Save Sets, the Unsuccessful Save Sets, the
Successful Save Sets, the Previously Successful savesets and the Cloned
Save Sets. Each client in the group will be listed in one or more of
sections categories (more than one if some save sets are in one cate‐
gory, and other save sets in another category). The clients are listed
in alphanumeric order, with the server listed last.
The header shows the name of the group and lists which clients failed,
unresolved, disabled and successful (with warnings). If the group was
aborted, the header includes an indicator of this as well. The header
also shows the time the group was started (or restarted, if the -R
option was used), and the time the savegrp completed. The failed
clients list in the header shows only those clients for which saves
were attempted, not those for which saves never started.
The Never Started Save Sets section is optional and is included only if
there are some save sets of some clients in the group that were never
started. This section can be seen when a savegrp is aborted, either by
killing the master savegrp daemon or by selecting the Stop function in
the Monitoring menu of NetWorker Management Console's Administration
window. Each entry listed in this section shows the client and save
set that was never started (or All if no save sets were saved for that
client). No other error messages should appear in this section.
The Unsuccessful Save Sets section shows all of the saves that were
attempted but failed. This section will only be present if at least
one save set failed. There are many reasons for a save to fail. The
most common are listed below. More reasons will be listed in the
future. It is important to differentiate between the many reasons for
a save to fail, so that the administrator can quickly determine the
cause and fix it.
Each entry in the Unsuccessful Save Sets section lists the client and
save set that failed, along with one or more lines of error and infor‐
mation messages. Each client is separated by a blank line, and all the
failed save sets for a client a listed together. Typical error or
information messages are listed at the end of this section, (without
the client:saveset prefix), with the necessary action(s) to take to
correct the problem.
Each entry in the Successful Save Sets section lists the client and
save set that succeeded, along with level of the save, the amount of
data saved, the time to run the save set, and the number of files
saved. Each entry may also be preceded by one or more warning or
informational messages, the most common of which are listed below.
These warning or informational messages are usually (but not always)
prefixed by ``* ''. A save set's output may include warnings; these do
not necessarily mean the save set was unsuccessful. The success
threshold attribute is used to determine if the warning(s) effect
whether the saveset is reported successful or failed. See nsr_group(8)
for the definitions of success threshold and its effect on reporting
success/failure of save sets.
Also see mminfo(8) for the definitions of a successful and unsuccessful
save sets.
The Cloned Save Sets section refers to the save sets cloned, and not
the clients that originated those save sets. The output shown in this
section is the output of the nsrclone command. See the nsrclone(8) man
page for information on the output of nsrclone.
The Previously Successful Save Sets section is optional and is included
only if the group was restarted and there were some savesets completed
in previous runs of the savegroup. This section is identical to Suc‐
cessful Save Sets section.
The following is a list of common informational, warning and error mes‐
sages found in the completion notification. This list is not complete.
The messages you see may vary slightly from those shown here due to
differences in the operating system vendor-supplied error messages.
Since many messages include client or server names, it is most effi‐
cient to look for a keyword in the error message. The messages are
listed below in alphabetical order, by the first non-variable word in
the message. (Note: initial words like "save", "asm" and "savefs" may
or may not vary, and initial pathnames are always assumed to vary).
aborted
This informational message only occurs when you abort a running
savegrp, generally by selecting Stop from the Monitoring menu of
NetWorker Management Console's Administration window. It means
that the specified save set had started saving, but had not com‐
pleted when the savegrp was aborted. The session (in the Moni‐
toring display of NetWorker Management Console's Administration
window) for this save set may not disappear immediately,
especially if savegrp's attempt to kill the save session fails.
The save set will be retried if and when you Restart the savegrp
(e.g. from the Groups tab of the Monitoring display).
Access violation from client - insecure port N
This message, generated by the save command on client, means
that save is not setuid root. Make sure that the save command
on the client is owned by root and has its setuid bit set. If
save is on an NFS mounted filesystem, make sure the filesystem
was not mounted on that client using the "-nosuid" option.
Access violation - unknown host: client
This message is caused when then the client's hostname and IP
address are not correctly listed in one or more of /etc/hosts,
NIS or DNS on the server. You must change the appropriate host
table (depending on which one(s) are in use on your server) to
list the client's name as it is known to NetWorker (client's
primary name), or you must add the name listed at the end of the
error message to the aliases attribute of the client's Client
resource(s).
asm: cannot open path: I/O error
This message generally means that there are bad blocks on the
disk(s) containing the specified file or directory. You should
immediately run a filesystem check on the named client filesys‐
tem and check your client's system error log. If there are bad
block, repair them if possible, or move the filesystem to a dif‐
ferent disk.
asm: cannot stat path: Stale NFS file handle
asm: cannot stat path: Missing file or filesystem
These informational messages (or variants of them for other
operating systems) mean that the when save attempted to test the
named directory (to determine if it was a different filesystem
from the one currently being saved), the filesystem was NFS
mounted, but the mount point was bad. While this message does
not affect the saved data, it does mean you have a network or
NFS problem between the specified client and one or more of its
fileservers. You may need to remount filesystems on the client,
or perhaps reboot it to correct the problem.
/path/nsrexecd: Can't make pipe
/path/nsrexecd: Can't fork
fork: No more processes
The specified client-side resource has been exceeded. There are
too many other services running on the client while savegrp is
running. Inspect the client and determine why it has run out of
resources. The client may need to be rebooted. You should also
consider re-scheduling any jobs automatically started on the
client (e.g. via cron(8)) that run while savegrp is running.
asm: chdir failed path: Permission denied
This message means that while backing up the specified save set,
save was unable to enter the named directory. This may mean
that save is not setuid root on the specified client, or that
the directory is an NFS mount point for which root is not
allowed access. Check the permissions on save on the specified
client (using ls(1)) and make sure that save is owned by root
and that the setuid bit is set.
connect to address AA.BB.CC.DD: message
Trying AA.BB.CC.DD...
These informational messages are displayed only when the -v
option is used. They mean that the connection to the client
failed on the address specified in the first line of the mes‐
sage. If the client has more than one IP address, savegrp has
attempted the address listed in the second line. Subsequent
lines of the completion mail show if this second address suc‐
ceeded. You may want to check and change your network routing
tables to avoid getting these messages.
Connection refused
This means the client machine is up, but it is not accepting new
network connections for nsrexecd (or rshd). This could mean the
client was in the process of booting when the savegrp attempted
to connect, or that the client had exceeded some resource limit,
and was not accepting any new connections. You should attempt
to log into the client and verify that it is accepting remote
connections. If the client is a non-UNIX machine, you may need
to start the NetWorker client on that machine. Refer to your
ClientPak installation for more information.
Connection timed out
This usually means the client has crashed or is hung. Make sure
the client has rebooted, and that nsrexecd is running on it (if
you are using nsrexecd). If the client is a non-UNIX machine,
you may need to ensure that the network protocols are loaded,
and that the NetWorker client is running on that machine. Refer
to your ClientPak installation for more information.
asm: external ASM `asm2' exited with code 1
This message generally accompanies another message reporting a
specific problem while saving a file or directory on the named
save set. The backup will attempt to continue and attempt to
save other data. Generally, the backup will not be listed in
the failed save sets section of the completion mail if any files
on the save set are saved successfully, even if it only saves
the top directory of the save set.
save: path file size changed!
This informational message is often generated when NetWorker
backs up log files. It may also occur for other files. For
files that you expect to grow while savegrp is running, you can
use a directive specifying that the logasm(8) should be used to
back up the file. See also nsr(5) and nsr_directive(5).
asm: getwd failed
This message means that while backing up the specified save set,
an attempt to determine the current directory's name failed.
This occurs on clients, generally running older versions of the
NetWorker ClientPak, on which the getwd(3) library call is bro‐
ken. You may want to contact EMC Tech Support to find out if
there is a patch available for your client platform to work
around this vendor-specific bug, or contact your operating sys‐
tem vendor to see if a more recent OS version addresses this
problem.
Group groupname aborted, savegroup is already running
This message is only delivered by itself. It occurs when the
named group has already been started or restarted (eg after a
reboot, or when requested via the Groups tab of NetWorker Man‐
agement Console's Administration window), either automatically
by nsrd(8) or manually, from the command line. You can use
ps(1) to find out the process id of a running savegrp. The
existence of a running group is determined by looking for a file
named /nsr/tmp/sg.group which, if existing and locked, means a
savegrp is running.
Aborting inactive job (id) client:saveset
The client has not sent any data to the server for the specified
inactivity timeout. savegrp will request nsrjob(8) to terminate
the backup in progress so that the hung client will not impede
other backups or cloning operations.
has been inactive for N minutes since time.
client:saveset is being abandoned by savegrp.
A backup of the specified save set started, but after N minutes
of no activity, savegrp gave up on the save set. Generally,
this means that the client is hung waiting for an NFS partition.
Unfortunately, NetWorker (or any other program) has no way of
reliably telling if an NFS partition will hang until after it
tries to access the partition. When the partition comes back on
line, the save will complete, despite that savegrp abandoned it.
You should check the client, since you sometimes may need to
reboot the client to unhang NFS partitions. Non-UNIX clients
also hang for other reasons such as defects in the operating
system implementation of their network protocols.
Host is unreachable
The NetWorker server cannot make TCP/IP connections to the
client. This generally means the network itself is not config‐
ured correctly; most commonly, one or more gateways or routers
are down, or the network routes were not set up correctly. You
should verify that the server can connect to the client, and if
not, check and, if necessary, reconfigure your routers, gateways
or routing tables.
Login incorrect
This message is generated when the remote user attribute for the
client is not set to a valid login on the client. Verify that
the remote user attribute for the client is set to the correct
login name. You may see this message even when running nsrexecd
if nsrexecd has not been started (or was killed) on the client.
asm: missing hard links not found:
This message is generated when a backed-up file had one or more
hard links that were not found. The message is followed by a
list of one or more file names which were backed up minus some
links. The message means that the files were either created
(with multiple hard links) while the backup was occurring, so
some of the links were missed due to the order of filesystem
tree walking, or the file (or some links) were removed while the
backup was occurring. Only those links that were found can be
recovered; additional links will have been lost. One can do an
additional incremental backup of the affected filesystem if a
consistent state for the affected file is essential.
lost connection to server, exiting
save: network error, server may be down
The backup of the named filesystem was begun, but the connection
to the NetWorker server closed part way through. This typically
means that the server machine rebooted, one or more NetWorker
server daemon processes were killed by the system administrator
or by the system itself (e.g. due to overwriting the binary or a
disk error in swap space), or there was some transport problem
that caused the network connection to dropped by the operating
system. Restart the save at a later time.
No save sets with this name were found in the media database;
performing a full backup
This informational message is added by savegrp to any save set
that is saved at the level full instead of the level found in
the client's schedule. Due to timing problems, you can occa‐
sionally see this message when the clocks on the client and
server are out of sync, or when savegrp starts before midnight
and ends after midnight. You may also get spurious messages of
this type from some versions of NetWorker client software back‐
ing up a NetWare BINDERY, which ignore the schedule and perform
a full save. In both these cases, the client re-checks the
level, and overrides the server's requested level.
No more processes
See "Can't make pipe" message information.
No 'NSR client' resource for client clienthostname
savefs: cannot retrieve client resources
This pair of messages occurs if the the client's hostname
changed (in /etc/hosts, NIS or DNS). You may also have deleted
the client's Client resource while savegrp was running. In the
former case, you will need to add the client's new name to the
aliases attribute of the client (this is a hidden attribute)
using nsradmin(8) (selecting the Hidden display option) or Net‐
Worker Management Console (double clicking on the client entry
from the Administration window's Configuration display). In the
latter case, no additional action is required if this deletion
was intentional (the next run of savegrp will not attempt to
save the client). If it was accidental, and you did not want to
delete the client, you should add the client back again and add
the client back into the appropriate group(s). The next time
savegrp runs, it will back up the client, just as if the client
had been down the previous day.
no output
The save set completed, but returned no status output. The most
common reasons are that the client crashed or lost its network
connection (i.e.. a router between the client and server
crashed) while the client was being backed up. Another is that
the disk on which the client status was being logged filled up
(perform a df /nsr/tmp to see if this was the case). To deter‐
mine if the save set was saved, you can use mminfo(8). For
example, run mminfo -v -c clientname -t '1 day ago' and look at
the flags column for the completion status. An 'a' flag means
it aborted. Use a more distant time (the -t option) to look
further back in time.
filesystem: No such file or directory
An explicit save set was named in the Client resource for the
specified client, and that save set does not exist (or is not
currently mounted) on the client. Make sure you spelled the
save set name correctly (and that it is capitalized correctly),
and log into the client and verify that the save set is mounted.
/path/nsrexecd: Couldn't look up address for your host
/path/nsrexecd: Host address mismatch for server
The nsrexecd daemon on the client managed to look up the server
in the client's host table, but the address listed there did not
match the address of the server. Every interface of the server
must have a unique name listed in the host table (possibly with
non-unique aliases or CNAME's), and each unique name must be
listed as a valid server to nsrexecd.
/path/nsrexecd: Host server cannot request command execution
/path/nsrexecd: Your host cannot request command execution
The server is not listed in nsrexecd's list of valid servers on
the specified client. The list of valid servers is either on
the nsrexecd command line (with one or more -s server options to
nsrexecd), or in a file (with the -f file option to nsrexecd).
If neither is specified, nsrexecd will look for a file named
servers in the same directory that contains the nsrdb configura‐
tion database (e.g. /nsr/res/nsrdb on a typical UNIX server).
Also the server may not be listed in one or more of /etc/hosts,
NIS, or DNS, on the client, in which case nsrexecd cannot vali‐
date the server until the client's host naming configuration is
fixed.
/path/nsrexecd: Invalid authenticator
/path/nsrexecd: Invalid command
These two messages should never occur in a savegroup completion
message. They mean that savegrp did not follow its protocol
correctly.
/path/nsrexecd: Permission denied
Permission denied
These similar messages are generated by nsrexecd and rshd,
respectively. In either case, the server does not have permis‐
sion to execute commands on the client. In the case of the
first message, make sure that the server is listed as a valid
server on the client (see "Host server cannot request command
execution", above, for details). In the case of the second mes‐
sage, which does not mention nsrexecd, make sure that "server‐
name" is listed in the client's /.rhosts file (or, if you have
set the remote user attribute for this client, the .rhosts file
in the home directory for that user on the client).
/path/savegrp: printing bootstrap information failed
See "unknown printer" message information.
reading log file failed
After the specified save set completed, savegrp was unable to
read the log file of the output status from the save set. This
generally means that someone, or an automated non-NetWorker
administrative program or script, removed the log file. This
message can also occur if the filesystem on which the client
logs are stored has run out of space (use df /nsr/tmp to deter‐
mine if this is the case). Verify that no scripts remove files
from /nsr/tmp (which is where savegrp stores the save set log
files).
request from machine server rejected
The server is not listed in the PC (NetWare or DOS) client's
list of acceptable servers. See your ClientPak installation
guide for instructions on adding the server to the client-side
list.
N retries attempted
1 retry attempted
One of these informational messages is prepended to a save set's
output if savegrp is unable to backup the data on the first try
and if the client retries attribute for the group has a value
greater than zero. In this case, the specified number of
retries was performed before the backup of the save set suc‐
ceeded or was finally marked as failed.
RPC error: details...
Cannot open save session with `server'
The save command generates this message if it is unable to back
up data to the NetWorker server. There are several possible
details. The most likely causes are: resources are exceeded on
the server so nsrd cannot accept new save sessions, nsrd actu‐
ally died since savegrp started (however, this is unlikely,
since you cannot normally receive a savegrp completion message
after nsrd dies, but you can see this when using the -p option),
there are numerous network errors occurring and save cannot open
a session to save its data (check this by running netstat -s and
see how many network errors are occurring; you may need to do
this several times a few minutes apart to get the change in
errors). Save cannot tell which of these three causes are the
real cause. If you see these errors frequently, and it looks
like a server resource problem, you might consider increasing
the value of the client retries attribute of the group resource
having these problems. This won't decrease the resource uti‐
lization, but will make savegrp more robust. (The trade-off is
that increasing client retries will increase the load on the
server even more).
nsrexecd on client is unavailable. Using rsh instead.
This informational message is only displayed when the -v flag
has been used for verbose information. This message means that
nsrexecd is not running on the client, and that savegrp is
attempting to use the rshd service instead for backward compati‐
bility with older versions of savegrp.
save: clientname2 is not on client's access list
This error occurs when the named client has more than one name,
for example, a short name, client, and a fully-qualified domain
name, client.legato.com. When the client attempts to connect
back to the NetWorker server to start a save, that client is
calling itself by the name client, which matches the client
resource name, but when the server looks up the client's network
address, it is getting back the name clientname2. If this is
correct, add the name clientname2 to the client's aliases
attribute and re-run the save.
save: path length of xxxx too long, directory not saved
This message can occur if you have a directory tree that is very
deep, or directory names that are very long. This message can
also occur if there are bad blocks in the specified filesystem,
or if the filesystem is corrupt. NetWorker limits the full
pathname to 1024 characters which is the system imposed maximum
on most systems. To save such directories, you need to rename
or move the directories so that the full pathname is shorter
than 1024 characters. If the filesystem appears to be corrupted
(for example, a very long pathname that looks like it has a loop
in the name), perform a filesystem check on the specified
client.
/path/save: Command not found
/path/savefs: Command not found
/path/save: Not found
/path/savefs: Not found
The save or savefs command could not be found in the specified
path. If you are using nsrexecd, this probably means that the
save or savefs command is not in the same directory in which
nsrexecd is installed (or that save or savefs was removed). If
you are using rshd for remote execution, then you need to set
the executable path attribute in the Client resource for this
client to be the directory in which the NetWorker executables
are installed on the client.
savefs: error starting save of filesystem
This informational message accompanies several other save or asm
messages listed here. This message means that savefs has
detected the failed save and has marked the save set as failed.
save: unknown host name: server
savefs: unknown host name: server
The host table on the specified client (either /etc/hosts, NIS
or DNS, depending on that client's configuration) does not
include the server's name. Add the server's hostname to the
specified client's host table. If you use DNS but the server's
Client resource name (i.e. the client resource for the server
itself) is not fully qualified (i.e. it looks like "server", not
"server.dom.ain"), and the server is in a different domain from
the client, add the name server to the domain table for the
domain containing the client. If you use NIS, this error means
that either the NIS hosts map does not contain the server, the
/etc/hosts file does not list the server, or the NIS master for
the specified client is otherwise mis-configured (the server is
a secondary server and there is no yppush(8) from the primary;
run ypwhich -m on the client to find which NIS server is provid‐
ing master translation).
savegrp: client rcmd(3) problem for command 'command'
This error message normally accompanies another, more specific,
error message. It is generated when the attempt to run the
specified command (usually save or savefs with several command
line parameters) failed on the specified save set. The previous
line of error output should include the more specific error
message (look for that message elsewhere in this section). Gen‐
erally, the problem is a bad hosttable configuration, or various
permissions denied errors (server not specified when starting
nsrexecd, or missing permissions in .rhosts if not using nsrex‐
ecd). If not, log into the NetWorker server as root and run the
command savegrp-p -v -c clientname groupname giving the appro‐
priate client for clientname and groupname . This verbose out‐
put should include the necessary additional information needed
for fixing the problem.
savegrp: suppressed N lines of verbose output
Sometimes a backup will generate a huge amount of output, such
as when one runs savegrp-v. When savegrp updates the group
completion attribute in the server, it may suppress some initial
lines of of this output, since logging all of the output to the
completion attribute can cause nsrd to use an unexpectedly large
amount of memory. The entire output of savegrp-v can be found
in the daemon.log.
savegrp: suppressed N lines of output -
check daemon.log for details.
The savegrp completion notification gets truncated if it is 1024
characters or longer. The daemon.log and NetWorker Management
Console will have details of the complete backup. If the
"NO_SUPPRESS" file is created in the /nsr/tmp directory, no sup‐
pression of the savegrp completion notification is done.
socket: All ports in use
The NetWorker server has run out of socket descriptors. This
means that you have exceeded the socket resource limit on your
server. To avoid such future messages, you should determine
what other network services are running while savegrp is run‐
ning, and consider re-scheduling either savegrp or the other
service(s). You can also reduce the parallelism in the nsr_ser‐
vice(5) resource, to reduce the resource utilization.
socket: protocol failure in circuit setup.
The client does not seem to support the TCP/IP protocol stack,
or has not used a privileged port for setting up its connection.
The latter could occur if you use nsrexecd but did not start it
as root on the specified client. The nsrexecd daemon must run
as root on each client.
path: This data set is in use and cannot be accessed at this time
This message is generated by save sets on PC clients running DOS
or NetWare. The NetWorker client software on these systems can‐
not back up files open for writing, due to the interface pro‐
vided by the operating system. This message actually comes from
Novell's TSA and is not changeable.
unknown host
The specified client is not listed in the host table on the
server (note: a similar "save" or "savefs" specific message is
described above). Depending on your host configuration, this
means the client is not listed in one (or more) of /etc/hosts,
NIS, or the Domain Name Service. If you use fully qualified
domain names, you may need to make a new client resource for
this client, using that fully qualified domain name (i.e. name
the client resource "mars.legato.com", not "mars").
printer: unknown printer
path/savegrp: printing bootstrap information failed
(reproduced below)
This message, or similar messages, accompanies the bootstrap
information when savegrp was unable to print the bootstrap on
the printer. You need to either specify a different printer in
the printer attribute for the group, or configure your print
server to recognize the printer (by default, your system's
default printer is used). The bootstrap information is listed
as part of the savegrp completion mail. You should print out
this information immediately, in case your server has a disaster
and loses a disk, and fix the printer name used by savegrp.
Warning - file `path' changed during save
This warning message is generated when save notices that the
file's modification time changed while the file was being backed
up. NetWorker does not attempt to lock files before saving
them, as this would make backups run extremely slowly. You may
wish to backup files which generate this message manually, to
ensure that a consistent copy is saved. NetWorker does not
attempt this automatically, to avoid trying forever on the same
file.
Warning: `client' is not in the hosts table!
This message is generated by a save or savefs command run on the
specified client to save that client's filesystems. The
client's hostname is not listed in the host table on the client
(either /etc/hosts, NIS or DNS, depending on that client's con‐
figuration). This almost always results in a failed save. Fix
the client's host table and re-run the save.
asm: path was not successfully saved
This message generally accompanies one or more other more-spe‐
cific messages for the save set. The specified path within the
current save set was not saved successfully. The backup will
continue trying to back up other files and directories on the
save set.
asm: xdr_op failed for path
This error can be caused by several possible conditions (for
example, out of memory, defective networking software in the
operating system, an external ASM unexpectedly exiting, a lost
network connection). If it was due to a lost network connec‐
tion, then the NetWorker server most likely exited (due to
nsr_shutdown). After restarting the server, rerun the group.
If due to an ASM exiting unexpectedly (in this case, the message
should be accompanied by a message describing which ASM exited
unexpectedly), you may have found a bad block on the disk, or
perhaps a defect. Check if the client ran out of memory (there
may be console messages), and verify that there are no bad
blocks on the save set's disk. If there were network errors,
there may also have been messages logged by other programs on
the system console (client or server), or to system log files.
FILES
/nsr/tmp/sg.group A lock file to keep multiple savegrps of
the same group from running simultaneously.
/nsr/tmp/sg.group.client.* Temporary files used to log the output of
individual save sets for the named group
and client.
/nsr/tmp/ggroup* On filesystems with short names (less than
64 characters), the temporary files used to
log the output of individual save sets for
the named group.
SEE ALSOls(1), ps(1), nsr_getdate(3), rcmd(3), fstab(5), nsr(5),
nsr_directive(5), nsr_notification(5), nsr_service(5), nsr_group(5),
nsr_schedule(5), nsr_resource(5), mminfo(8), nsrssc(8), netstat(8),
nsr(8), nsradmin(8), nsrjobd(8), nsrexecd(8), nsrwatch(8), rshd(8),
save(8), savefs(8), pathownerignore(5), yppush(8).
NetWorker 7.3.2 Aug 23, 06 SAVEGRP(8)
