Once every day or two we see a series of warning in the logs about snapshot cleanup, and I was wondering if this is something we should be concerned about, or if there is something we can do to prevent it. We are using version 7.1.4. Sample message:
I’m thinking this may be related to the daily incremental backup jobs we have scheduled. Fun fact, my incremental backups are nearly as large as my full/init backup!
Hi carter,
The first trace is expected if there is a long running transaction of some kind. Its possible the backup is such a transaction. You can ignore that trace for the most part, it just means there are files on disk MemSQL could cleanup, but its blocked from doing so until the long running transaction that maybe using those files completes.
As for the incremental backup file sizes your seeing, the files your looking at there are for rowstore data. We don’t support incremental backup for those yet. Every incremental backup does a full backup for rowstore data and then copies incremental delta for columnstore data.
-Adam
Thanks Adam. However I’m still confused on the incremental backup question. 95% of our data is in columnstore tables, and the data isn’t changing much, our differentials should be small, correct?
You are correct, your differentials should be relatively small.
Try doing a du -sh on your incremental backup target directory before and after the next backup.
- Nate
Hi Nate,
here is the before (after a full) and after (after 2 incrementals). Incremental definitely does not seem to be helping much! Thoughts?
Is this an updated cluster from pre 7.0?
Also another experiment that we can try is taking an incremental immediately after taking the backup, this should be the best case scenario with the incremental backup.
Hi,
The cluster is a clean 7.0 installation, however the database was restored from a 6.7 backup. Could that be the issue?
Thanks,
~Steve