[MooseFS-Users] Follower temporary in desync state

[Casper L. <cas...@pr...>]

Hi All,

I have created a script to monitor my MooseFS masters with Nagios, and I
find that quite some times the follower master is in desync state.

On a check interval of 5 minutes, a desync state happens about a dozen
times a day. The amount of time time happens seems to be correlating with
the (write?) load the file system is taking. I'm guessing this is expected
behaviour?

I'm reading the MATOCL_INFO (PROTO_BASE+511) data, the same mfs.cgi does in
line 1836:
   *workingstate,nextstate,stablestate,sync,leaderip,changetime,metaversion
= struct.unpack(">BBBBLLQ",data[101:121])*

*sync* seems to be 1 most of the time, but sometimes 0. *metaversion* is
updated continuously, but *changetime* is always the same. On my system it
is a timestamp set at the date and time my master started. Should
*changetime* be the date and time of *metaversion*? i.e., continuously
updating on any file system changes? If that is the case I can change my
nagios script to only warn me if the follower is a certain number of
seconds behind.

Is there a better way to monitor follower master status?

Greetings, Casper