Unsolved
This post is more than 5 years old
22 Posts
0
2211
March 2nd, 2010 06:00
RLP luns are not being released
We have a CX-240 with about 5 TB of storage. Our OS is aix and we are running a semi-large Oracle database (550G). Our RLP is 400Gb of 5Gb luns spread across the array.
I've got a rolling snapshot setup to take a snap of the Oracle database every 6 hours and I'm retaining 5 copies. For about 8 months now that has been working fine and our RLP has been about 50% used. Over the last 3 or 4 weeks 4 of the source luns has started using up all the RLP in sight. Our RLP is 95% full now. My concern is that the luns are not being recycled as the snapshots are aged out. Not real sure what could be causing this and we're fairly new to this so I'm also not sure if there is some kind of management command we are supposed to be issuing to clean up these luns and make sure they are getting re-used.
Any guidance is appreciated.
MohammedMansoor
21 Posts
0
April 9th, 2010 03:00
If the reserved LUN fills up and the SP's LUN pool has no additional LUNs, the software automatically terminates the session that is trying to allocate reserved LUN space, logs an error, releases the reserved LUN(s) used by this session, and returns them to the SP's LUN pool. The software also destroys all copy-on-first-write data stored in the reserved LUN pool for that session. At this point, the snapshot becomes inactive and any host that has mounted volumes on the snapshot will receive I/O errors and lose access. If you have multiple sessions of a single source LUN and the reserved LUN fills up and no additional LUNs available, when the production host modifies a chunk on the source LUN, resulting in a copy-on-first-write, every session that has the same chunk will be terminated. Other sessions that did not have this chunk will continue to run and use the reserved LUN space that the terminated sessions were using.
Fix: Each SP has its own reserved LUN pool, and before starting a session, the reserved LUN pool must contain at least one LUN for each source LUN that will be starting a session. You can add any LUNs that are available to either SP’s reserved LUN pool. Each SP manages its own LUN pool and assigns a separate reserved LUN (or multiple LUNs) to each SnapView source LUN. Multiple sessions of a single source LUN will share the same reserved LUN or LUNs.
This information along with other SnapView setup information can be found in the SnapView Administrators Guide.
Mohammed Mansoor
MohammedMansoor
21 Posts
1
April 9th, 2010 05:00
Once the sessions are all terminated , the RLP will be automatically freed
Ensure to terminate all session to free up the RLP
If you are sill facing issues please log a support Service Request
Mohammed Mansoor
BenA4
22 Posts
0
April 9th, 2010 06:00
what exactly do you mean by "terminate session" ?
I have it setup to run automatically via RM. The snap shots take place and the jobs show as complete. Is there something else that has to manually be performed?
In this instance, I deleted the five snapshots for the application set and that freed up all the reserved luns. This job has since then only aquired 17 RLP luns and our RLP is only about 30% used.
BenA4
22 Posts
0
April 12th, 2010 05:00
There are no errors.
Would a lun ever be released back to the pool without deleting all the snapshots associated with a job?
I believe what caused the huge growth of usage of luns was a large archival process that we ran. Typically the activity isn't that high and we were hoping after the activity had gotten back to normal the luns that were grabbed would be released. I didn't know if that was possible or not though as the blocks would probably be scattered all over all the luns.
MohammedMansoor
21 Posts
0
April 12th, 2010 05:00
Thats the way it should work ... are you getting any errors in the Replication Manager Job History , if yes can you provide the same for further support ??