Gridstor using Shared Disk library with SAN mount path
Hi everyone,
We are in the progress of testing how to setup Gridstor with shared library which mount paths are assigned to each Media Agent locally (no NAS available).
Assuming we have Disk Library configured with MA01 and MA02, mount paths are shared to each other (read access only) configured as below:
Disk Library
MA01
MA02
Mount Path A
Mount Path C
Mount Path B
Mount Path D
\\Mount Path C (Read Only)
\\Mount Path A (Read Only)
\\Mount Path D (Read Only)
\\Mount Path B (Read Only)
However, if MA01 down, with Gridstor backup would continue to run on MA02.
But when comes to restore (coincidently), MA02 would not be able to access shared mount path on MA01.
We being told that we could re-zone/mount the Mount Path A and Mount Path A to MA02, configure mount path as part of MA02 then we could perform data restoration. Example as illustrated below:
Disk Library
MA01
MA02
Mount Path A
Mount Path C
Mount Path B
Mount Path D
Mount Path A (Read access only)
Mount Path B (Read access only)
When MA01 back online. We would like to unmount the Mount A and B from MA02, and rezone/mount back to MA01 as initial configured state and let backups and restore to work as normal.
Disk Library
MA01
MA02
Mount Path A
Mount Path C
Mount Path B
Mount Path D
\\Mount Path C (Read Only)
\\Mount Path A (Read access only)
\\Mount Path D (Read Only)
\\Mount Path B (Read access only)
Below is the drafted scenario:
1. MA01 offline 2. Mount Path A and Mount Path B present online on MA02. 3. Configure Mount Path setting - Mount Path A and B to be part of MA02. Enable Read access only. 4. Perform restore from jobs written to Mount Path A and B using MA02. 5. Verify jobs are restorable 6. Unmount Mount Path A and Mount Path B from MA02. 7. Reconfigure Mount Path setting on MA02 8. MA01 online 9. Reconfigure Mount Path setting on MA01 10. Ensure backup jobs able to run on both MA01 and MA02
But we are not so sure how to play around with the disk library and mount path setting in console for step 3, 7, and 9 above.
Appreciate if any of you could share some experience and knowledge to achieve the scenario illustrated above.
Page 1 / 1
I don’t have anything valuable to add, except that if you have the resources clustering those media agents together is a significantly better solution than this sort of gridstor configuration.
I don’t have anything valuable to add, except that if you have the resources clustering those media agents together is a significantly better solution than this sort of gridstor configuration.
Hi @christopherlecky ,
Thanks for reply, we are currently have all MAs running on GFS2, which CommVault stated
GFS2 is not supported for use as a disk library target due to locking / performance issues that do not scale.
Thus we are exploring setting up Gridstor with available resources, and we need to get some firm info on how to play around with the setting given the limitation of local SAN mount path presented on each MA.
Thanks and regards
Hi @SMSSB
This can be done and I can explain how to do it but would need to know how the type of storage and how it's being presented to each MA. Is it iSCSI, local disk, etc?
Hi @Matt Medvedeff ,
Thanks for reading this post.
Storages are all SAN FC connected.
Moving out from GFS2, Volumes are going to present to each MA locally via FC connection, we will need to configure 4-node partitioned dedupe and using Gristor for the group of MAs.
We will skip Shared SAN Data Server setup as this is will add more complexity to the environment. Appreciate if you could share some ‘gotcha’ or steps/caution that I should be aware of, especially for step 3, 7, and 9 above.
Thanks and regards,
Bump, following up if anyone could share some ideas on this.
Thanks
bump, still searching for info on this matter
@SMSSB , I know @Matt Medvedeff was looking into this for you. I’ll ping him on the side.
@SMSSB , I know @Matt Medvedeff was looking into this for you. I’ll ping him on the side.
Hi @Mike Struening ,
Tq for the info, while waiting for Matt Medvedeff and others, we have actually running some test with 2-node set-up as mentioned above, backup was not successful due to ddb thingy. Case Incident 220601-26 if any of you able to read somewhere.
Thanks and regards
Apologies for the delay on this. So when MA01 is offline and the storage is attached to MA02:
You would go into the properties of the mount paths in red → Sharing Tab → Edit → Change the Details here from the Network path to the Local path of the Disk now locally attached to MA02
Repeat the process for both Disks.
Make sure to note the exact configuration and paths listed in the Mount path details page as the “gotcha” here is to make sure you’re maintaining the directory structure when the configuration is changed from Network to Local paths.
When MA01 is back online, you would just re-edit these details and point the disks back from Local to Network.
Make sense?
Apologies for the delay on this. So when MA01 is offline and the storage is attached to MA02:
You would go into the properties of the mount paths in red → Sharing Tab → Edit → Change the Details here from the Network path to the Local path of the Disk now locally attached to MA02
Repeat the process for both Disks.
Make sure to note the exact configuration and paths listed in the Mount path details page as the “gotcha” here is to make sure you’re maintaining the directory structure when the configuration is changed from Network to Local paths.
When MA01 is back online, you would just re-edit these details and point the disks back from Local to Network.
Make sense?
Hi @Matt Medvedeff ,
Thank you for enlighten me with the info. This should be workable on Windows Partitioned Dedupe. Do you mind share how do we deal with Linux MA? As the mount paths are configured as local path.
as mentioned in BOL.
Note: If the MediaAgent is a UNIX/Linux MediaAgent, then you must select the Local Path option. To access an Network File System (NFS) share, you must mount the share to your UNIX/Linux MediaAgent. Once the share is mounted, it appears as another local directory on the system. You can provide this directory location in the local path. Configure the NFS mount option local_lock to none. The Commvault software does not support local_lock option set to all, posix or flock, as these settings can cause data inconsistency. If you apply any of these configurations, the mount paths may not come online.
Scene 1
Linux MA01 (Offline)
Linux MA02
\Mount Path A
\Mount Path C
\Mount Path B
\Mount Path D
Scene 2
Linux MA01 (Offline)
Linux MA02
\Mount Path A
\Mount Path C
\Mount Path B
\Mount Path D
\Mount Path A
\Mount Path B
I have some doubts for linux MA.
Since we can only configure as local path, when MA01 down, we configure \Mount Path A & B on MA02 for restore purpose only. Would it cause the Mount Path corruption/ conflict with the media library id/label?
When MA01 back online, shall we remove the \Mount Path A & B configuration from MA02? You mentioned earlier we would just need to edit sharing from local to Network share for WIndows. How about for Linux MA?
Thanks and regards
Hello mate, how's it going?
I am writing to ask you how you finally solved the Linux MA scenario of shared data paths, did you identify any possibility of having HA to be able to restore if an MA goes down?
Thank you in advance for your attention.
Best regards
Hello All
Is there any update for Liunx MA mouth paths shares as read only
