Under what conditions are ReadIndex’es rebuilt on startup?
We got one customer, with very slow virtual disks, where it happens on every startup, and takes about 10 min (and will take longer as it grows). It’s a bit of a problem, since their it/ops are very quick to reboot servers regardless of issue…
We haven’t been able to disable write-caching (just get ‘not supported’-message), is this related, some kind of dirty shutdown?
/Peter
No this is normal basically the last up to 1m are kept in memory index then batched 1m at a time down to disk. On a restart we replay up to 1m. This size (index memory size) is configurable. Use a lower value and less will be replayed on startup. Knocking it down to 100k would make your slowest startup about 1m I guess
OK, thanks, guess it’s fast enough not to be noticed on most systems.
Is it the MaxMemTableSize option?
Yep! It will cause smaller mem tables to be used which means less data to load at start up.