3 Dec
2015
3 Dec
'15
10:22 p.m.
Philip Gardner, Jr. wrote:
During the replication events, we are seeing the controller hosting this particular datastore basically getting crushed and issuing B and b CPs. Here is some output of sysstat during one of the replication events:
Many/most ppl here prob already know this, but just to be ultra safe: there's no difference in practice w.r.t. badness between a 'B' (back-2-back CP) and a 'b' (deferred back-2-back CP).
Exacly what triggers a 'b' vs a 'B' I don't know.
It's not that the distinction is unimportant, it is. Just they're equally bad for performance
/M