Question

Is ceph supported as mount path of a dedup disklib

  • 23 December 2022
  • 3 replies
  • 115 views

Badge +4

My customer has configured a dedup disklib. The shared mount path is a directory in a filesystem working on ceph layer. One night he found a lot of jobs waiting because the library was open for jobs but did not write data to disk. There was an unknown error on the ceph layer and a reboot of nodes solve this issue.  The point is that the library was online and did not went offline during the “ceph error”. Customer reach the maximum limit of jobs and new partial important backup jobs could not start. Now we analyzed what happened. 

My question here:Is ceph supported as target of a disklib.

In BOL I found entries about kubernetes, S3 connection etc. but nothing about using it as mountpath in a filesystem. I know that ceph supports block, file and object level.

Thanks in advance for your answers 

Joerg


3 replies

Userlevel 4
Badge +10

Hi @Joerg 

We Fully support Ceph configured as S3 compatible Storage configured as a Cloud Library in Commvault -https://documentation.commvault.com/2022e/essential/9236_supported_cloud_storage_products.html

Just to confirm, your customer mounted presented and mounted the storage to the Media Agent and configured it as a Disk Library (not cloud) in CV?

Userlevel 7
Badge +19

Hi @Joerg 

We Fully support Ceph configured as S3 compatible Storage configured as a Cloud Library in Commvault -https://documentation.commvault.com/2022e/essential/9236_supported_cloud_storage_products.html

Just to confirm, your customer mounted presented and mounted the storage to the Media Agent and configured it as a Disk Library (not cloud) in CV?

Well I can clearly read from the information that @Joerg refers to a shared mount path based on a fileshare. 

@Joerg I think it also really depends what the actual issue was and how Commvault detected it. It seems Ceph was partially/half broken. This would mean Commvault was still able to read the data from the library? It might be that Commvault in this case was not able to peperolie detect that the disk library wasn't working correctly. 

Now there are a few things of course to look into like for example what was Ceph itself telling/reporting? It was clearly an issue with the storage, so I would assume it should have reported the issue. Other suggestion I would consider implementing is the alert that allows you send an alert when there are too many pending jobs in the job queue. This way Commvault could have alerted you that something out of the ordinary was going on. 

Userlevel 7
Badge +23

In general ‘yes’ - clustered type file systems are supported for disk libraries:

https://documentation.commvault.com/2022e/expert/9320_clustered_file_system_storage_support.html

However, my personal opinion is that it is not a good idea unless you are an expert with that file system. Commvault has a non-standard read/write profile compared with typical application like corporate fileshares, which can often mean the filesystem has to be tuned or altered to accommodate the load and achieve acceptable performance. You can see that GPFS2 is not allowed as it became a complete disaster, being unable to handle the type of load profile we had.

While Commvault can generally write to anything, we can’t help to fix the underlying layers if things go wrong or don’t work properly.

 

Reply