How do engineers deal with RAID controller battery "relearn" cycles?
As noted in: What's a "battery relearn" on a LSI MegaRaid?, relearn cycle discharges the RAID controller battery (BBWC or BBU), thus removing the write cache acceleration. The battery's life is checked and once charged, the write cache is reenabled. This has an obvious impact on server I/O performance for the duration of the relearn cycle. I think this occurs monthly.
The performance degradation has been noted, especially on database systems:
My background is in HP ProLiant servers, whose Smart Array controllers do not go through this exercise (or at least have more proactive battery life monitoring). This seems to be a terrible feature (maximum inconvenience, little gain), but I'm in an environment with many LSI controllers (on Supermicro hardware) and would like to see if a blanket policy can be applied to the relevant systems.
- What is the default schedule of the relearn cycle on an LSI controller?
- Are these relearn cycles useful?
- Should this feature be disabled?
- If you choose to leave this feature enabled in your environment, how do you handle scheduling? Do you schedule this manually or allow the controller to set its own schedule?
- Are Dell Perc controllers affected in the same manner? (LSI is the OEM)