Restore and mounts
I am doing a file system restore. The number of volumes for this node is 35 and is collocated by filespace. In the last 24 hours there has been 700 tape mounts for this restore session. One volume has been mounted 346 times. Total amount restored is about 200 GB. q ses f=d tells me that this is a NoQueryRestore. Is this to be expected? Regards Hans Chr.
Re: Restore and mounts
This could be normal if TSM is trying to recreate all the directory structures. It creates this first, before restoring actual data. With the newer versions of TSM, using a directory class management (DIRMC) shouldn’t be necessary, since ACL information is applied at a later point in time. However with that said, I’ve seen fileservers with millions of directory structures that could be spread across many tapes, or even one tape. You may want to open a ticket with support for confirmation, but the symptoms you are reporting are similar to a problem I had a while back. See this technote with a bit more background. http://www-01.ibm.com/support/docview.wss?uid=swg21669468 On Wed, Dec 17, 2014 at 3:37 AM, Hans Christian Riksheim wrote: > > I am doing a file system restore. The number of volumes for this node is 35 > and is collocated by filespace. > > In the last 24 hours there has been 700 tape mounts for this restore > session. One volume has been mounted 346 times. Total amount restored is > about 200 GB. > > q ses f=d tells me that this is a NoQueryRestore. > > > Is this to be expected? > > > Regards > > Hans Chr. >
Re: Restore and mounts
Thanks, Nick. Of course this was the one TSM server where I forgot to create the DIRMC diskpool and that explains the restore behavior. Regards, Hans Chr. On Wed, Dec 17, 2014 at 2:52 PM, Nick Marouf wrote: > > This could be normal if TSM is trying to recreate all the directory > structures. It creates this first, before restoring actual data. > > > > With the newer versions of TSM, using a directory class management (DIRMC) > shouldn’t be necessary, since ACL information is applied at a later point > in time. However with that said, I’ve seen fileservers with millions of > directory structures that could be spread across many tapes, or even one > tape. > > > > You may want to open a ticket with support for > confirmation, but the symptoms you are reporting are similar to a problem I > had a while back. > > > > See this technote with a bit more background. > > > > http://www-01.ibm.com/support/docview.wss?uid=swg21669468 > > > > On Wed, Dec 17, 2014 at 3:37 AM, Hans Christian Riksheim < > bull...@gmail.com> > wrote: > > > > I am doing a file system restore. The number of volumes for this node is > 35 > > and is collocated by filespace. > > > > In the last 24 hours there has been 700 tape mounts for this restore > > session. One volume has been mounted 346 times. Total amount restored is > > about 200 GB. > > > > q ses f=d tells me that this is a NoQueryRestore. > > > > > > Is this to be expected? > > > > > > Regards > > > > Hans Chr. > > >
Re: Restore and mounts
The point of that technote is that you do not need DIRMC anymore. On Wed, 17 Dec 2014 16:17 Hans Christian Riksheim wrote: > Thanks, Nick. > > Of course this was the one TSM server where I forgot to create the DIRMC > diskpool and that explains the restore behavior. > > Regards, > > Hans Chr. > > On Wed, Dec 17, 2014 at 2:52 PM, Nick Marouf wrote: > > > > This could be normal if TSM is trying to recreate all the directory > > structures. It creates this first, before restoring actual data. > > > > > > > > With the newer versions of TSM, using a directory class management > (DIRMC) > > shouldn’t be necessary, since ACL information is applied at a later point > > in time. However with that said, I’ve seen fileservers with millions of > > directory structures that could be spread across many tapes, or even one > > tape. > > > > > > > > You may want to open a ticket with support for > > confirmation, but the symptoms you are reporting are similar to a > problem I > > had a while back. > > > > > > > > See this technote with a bit more background. > > > > > > > > http://www-01.ibm.com/support/docview.wss?uid=swg21669468 > > > > > > > > On Wed, Dec 17, 2014 at 3:37 AM, Hans Christian Riksheim < > > bull...@gmail.com> > > wrote: > > > > > > I am doing a file system restore. The number of volumes for this node > is > > 35 > > > and is collocated by filespace. > > > > > > In the last 24 hours there has been 700 tape mounts for this restore > > > session. One volume has been mounted 346 times. Total amount restored > is > > > about 200 GB. > > > > > > q ses f=d tells me that this is a NoQueryRestore. > > > > > > > > > Is this to be expected? > > > > > > > > > Regards > > > > > > Hans Chr. > > > > > >
Re: Restore and mounts
Interesting technote. First the DIRMC is not needed. "Therefore the original reason to cache directories on disk no longer applies." Then it is useful after all. "Nevertheless, the DIRMC option is still useful. If you do not specify this option to associate a management class with directories, the client, during backup, uses the management class in the active policy set of your policy domain with the longest retention period, which could point to a storage pool on tape. This might result in unwanted mount requests during backup/restore." But DIRMC can cause problems with backups. "Having the DIRMC point to a disk storage pool separate from the backup objects might cause the symptom as described with technote swg21247892 (Unexpected high number of EndTxn processing during back up, see link section below)" This is fun! Regards, Hans Chr. On Wed, Dec 17, 2014 at 6:06 PM, Steven Langdale wrote: > > The point of that technote is that you do not need DIRMC anymore. > > On Wed, 17 Dec 2014 16:17 Hans Christian Riksheim > wrote: > > > Thanks, Nick. > > > > Of course this was the one TSM server where I forgot to create the DIRMC > > diskpool and that explains the restore behavior. > > > > Regards, > > > > Hans Chr. > > > > On Wed, Dec 17, 2014 at 2:52 PM, Nick Marouf wrote: > > > > > > This could be normal if TSM is trying to recreate all the directory > > > structures. It creates this first, before restoring actual data. > > > > > > > > > > > > With the newer versions of TSM, using a directory class management > > (DIRMC) > > > shouldn’t be necessary, since ACL information is applied at a later > point > > > in time. However with that said, I’ve seen fileservers with millions of > > > directory structures that could be spread across many tapes, or even > one > > > tape. > > > > > > > > > > > > You may want to open a ticket with support for > > > confirmation, but the symptoms you are reporting are similar to a > > problem I > > > had a while back. > > > > > > > > > > > > See this technote with a bit more background. > > > > > > > > > > > > http://www-01.ibm.com/support/docview.wss?uid=swg21669468 > > > > > > > > > > > > On Wed, Dec 17, 2014 at 3:37 AM, Hans Christian Riksheim < > > > bull...@gmail.com> > > > wrote: > > > > > > > > I am doing a file system restore. The number of volumes for this node > > is > > > 35 > > > > and is collocated by filespace. > > > > > > > > In the last 24 hours there has been 700 tape mounts for this restore > > > > session. One volume has been mounted 346 times. Total amount restored > > is > > > > about 200 GB. > > > > > > > > q ses f=d tells me that this is a NoQueryRestore. > > > > > > > > > > > > Is this to be expected? > > > > > > > > > > > > Regards > > > > > > > > Hans Chr. > > > > > > > > > >