ldmwndletsm
ADSM.ORG Senior Member
- Joined
- Oct 30, 2019
- Messages
- 232
- Reaction score
- 5
- Points
- 0
Hi, Kind of a weird situation here. I need to restore a large file system that contains a number of hard links under sundry subdirectories. This restore is from tape. My past experience with this is that the resource utilization MUST be set to 1. Otherwise, the results are unpredictable when restoring hard links wherein some may not be recreated. Otherwise, if set to 1 then all is well. The following IBM article supports this:
The problem is that this restore cannot be done in just a few hours. There are a huge number of inodes. Even if I start the restore early, it will take all day and into the evening, possibly finishing late at night. There are a lot of file systems on this client, with some having a large number of inodes, so the resourceutilization is set to 10 (dsm.sys) to maximize parallelism, giving us 4 producer threads (sessions) and 4 consumer threads:
This works nicely, BUT I don't want to lower it to 1 since this would impact backup performance, and again, I can't complete the restore before it will be time to run another backup. Therefore, I was thinking that maybe I could restore the data to another node B (using grant proxynode) that has only a small number of file systems on it, and set its resourceutilization to 1. I have a non-local file system on node B (NFS mounted) where I can restore the data, and this is not backed up.
1. Will that work? Or will TSM insist that the resourceutilization needs to be set to 1 on the node where the data was originally backed up?
Based on the table in the above IBM link, setting it to 1 would yield a maximum number of sessions=1 and producer sessions=0, thus consumer sessions would be 1-0=1. So we'd have 1 consumer session and 0 producer sessions.
2. How would we be able to backup any data on the agent node B if the producer sessions is 0?
Granted, the number of file systems on the agent node is small, and they're not very big, so it doesn't take it long to traverse them, but if the restore overlaps with the backups (very probable) then how could the backups for node B even run if the producer sessions is 0? Could it?
3. Is the resourceutilization value used for both backups and restores?
4. Can the resourceutilization for the restore be set separately from the one used for backups (dsm.sys)?
IT02889: DOCUMENTATION ON HARDLINK RESTORE WHEN RESOURCEUTILIZATION GREATER THAN 1 IS MISSING LIMITATION
The limitation described in APAR IC57371 still exists in recent Tivoli Storage Manager versions.
www.ibm.com
The problem is that this restore cannot be done in just a few hours. There are a huge number of inodes. Even if I start the restore early, it will take all day and into the evening, possibly finishing late at night. There are a lot of file systems on this client, with some having a large number of inodes, so the resourceutilization is set to 10 (dsm.sys) to maximize parallelism, giving us 4 producer threads (sessions) and 4 consumer threads:
Optimizing the number of multiple sessions to run
IBM Spectrum Protect clients can establish concurrent sessions to back up and restore data. The creation of concurrent sessions is controlled by an algorithm within the client software; you cannot directly control this algorithm. The default behavior is to use two sessions: one to query the...
www.ibm.com
This works nicely, BUT I don't want to lower it to 1 since this would impact backup performance, and again, I can't complete the restore before it will be time to run another backup. Therefore, I was thinking that maybe I could restore the data to another node B (using grant proxynode) that has only a small number of file systems on it, and set its resourceutilization to 1. I have a non-local file system on node B (NFS mounted) where I can restore the data, and this is not backed up.
1. Will that work? Or will TSM insist that the resourceutilization needs to be set to 1 on the node where the data was originally backed up?
Based on the table in the above IBM link, setting it to 1 would yield a maximum number of sessions=1 and producer sessions=0, thus consumer sessions would be 1-0=1. So we'd have 1 consumer session and 0 producer sessions.
2. How would we be able to backup any data on the agent node B if the producer sessions is 0?
Granted, the number of file systems on the agent node is small, and they're not very big, so it doesn't take it long to traverse them, but if the restore overlaps with the backups (very probable) then how could the backups for node B even run if the producer sessions is 0? Could it?
3. Is the resourceutilization value used for both backups and restores?
4. Can the resourceutilization for the restore be set separately from the one used for backups (dsm.sys)?