What settings do people typically use for their MGM namespace compactification in production?
Under Beryl-Aquamarine, we were using
ns compact <run at about 1am> 86400 all-repair, and this worked well for us.
Under Citrine, the behaviour is a little different, and we’re after comments from others what they’re using.
Still on aquamarine and I simply occasionally do:
ns compact 60 0 all
I would love to be able to do one regularly. Is this advised? I’ve had hangs during compacts and don’t want to push my luck. Also, is the -repair option recommended for regular use or only when things are awry?
We ran daily compactifications on Beryl-Aquamarine for years without issue, so long as we always had enough disk space.
Expect things to go really wonky if you run out of disk space during compacting. But we never had hangs during compacting, things seemed to work fine.
We had an unusual issue last night (thanks to @luca.mascetti for working through it with us) where under Citrine, a slave didn’t load the new files correctly into memory, and bugged out. For write access to resume to the master MGM, we needed to shut down the slave MGM.
Citrine appears to do compacting quite differently, and we identified some differences last night we’ll have to submit a ticket for clarity on.
I just identified with Luca a problem, which would keep clients away while the compaction is running for all ‘open’ calls. I will commit the fix and we will make a tag for CITRINE today.