On Sun, 9 Oct 2016, Songbo Wang wrote: > Hi, Sage, Sam > > I got a issue about the parameter 'paxos_propose_interval' with the > default value 1.0. > > I think this parameter aim to reduce the Monitor pressure to call the > propose_pending(). But, this introduces another problem, when a osd is > started, it will send 3 kinds of requests to Monitor (e.g. MOSDBoot, > MOSDPGTemp, MOSDAlive). Without other changes, It will take 3 seconds > at least. which will increase the time that blocks the whole cluster. > > If the parameter 'paxos_propose_interval' is set to 0.05 or smaller, > it loses the propse it designed. > > I think we can introduce some special process, for some kinds of > messages (e.g. MOSDPGTemp, MOSDAlive), to process them immediately. Currently we essentially have two classes of messages: those that are processed immediately (no delay), and those that respect the paxos_propose_interval. We can either introduce a second tunable and create a third class (process quickly) with something like .05, or we can say that all of the current 'no delay' messages should delay for a short period (like .05). Any preferences? sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
