[Sheepdog] How to know if sheepdog is recovering chunks
sirio81 at gmail.com
Thu Sep 1 09:54:12 CEST 2011
During my speedtests, I removed from the cluster a node and inserted
it again several times.
My cluster has 3 nodes and --copies=2.
When node3 was down I could see network and disks activity from atop but
collie cluster info
doesn't give any clue if there is a sync/rebuild on the way.
In a previous discussion we said it's not a good idea to insert back a
node while a syncronization is on.
1) is there a way to be sure no sync are woking?
Looking at atop during the sync after node3 is gone off, I noticed
that 'si' and 'so' (byte in and byte out the netwrok card) were most
of the time 10Mbps and 10Mbps.
Sometime they were up to 50si 10so or viceversa.
Disks were realy not so busy (I don't remeber now but less than 50%
read or write sure).
Node1 and node2 are connected by gigabit link so I was aspecting to
see much more activity.
>From the other side, if the sync process use too much resources, you
mught ght I/O error on the guests.
2) does sheepdog limit the sync speed? If yes, is it possible the set
the sync speed?
e.g. I know drbd allow to set the sync speed limit.
More information about the sheepdog