"cbstats ... all" returns stats for the entire cluster, not individual nodes
As far as I can tell, cbstats is returning stats for the cluster as a whole instead of the individual node stats. So for things like cmd_get, if I connect to one of the two nodes in a cluster (that has 1 replica), and do 1000 gets, the cmd_get will go up by 1000 on each of my two nodes. Also, things like uptime are the same on the two nodes, even though one of the servers was rebooted. I'm not sure I see the benefit of this.
Is there some alternative way to get per-node stats? Using the "stats" command while telneting to the bucket appears to give the same results as cbstats.
Here is an example of what tipped me off...
node22$ /opt/couchbase/bin/cbstats node22:11909 all | grep uptime ; /opt/couchbase/bin/cbstats node23:11909 all | grep uptime uptime: 344424 uptime: 344424 node22$ uptime 19:16:58 up 2 days, 1:50, 1 user, load average: 0.04, 0.16, 0.19
I can pull some more stats together later, if this isn't enough to show an issue.
Can you go through port 11210 and use the -b parameter to specify the bucket name?
That did the trick. Thanks mikew.
I just want to note the reason here for why this happens.
When you send a stats command through moxi the response might be aggregated (and this might actually happen incorrectly). In any case you shouldn't ever send a stats command through moxi since this isn't something we actually support. Hopefully at some point we will add something that makes this stop working so that people don't get confused.
The cbstats command should only be getting stats from a single node. Also, from a node perspective none of te nodes know about each other so it isn't possible for them to exchange stats. Can you post the stats here so I can look at them?