Our ZKDashboard has been failing to read the status of the leader the
past few weekends after cluster reboot. This seems to be because our
reboot schedule is not causing an even distribution of sessions on
restart and so the leader has a huge number of sessions, and this
seems to cause the stat command to truncate the output when executed
from a different data center than the leader. Eg, normally it should
look like

stat
Zookeeper version blah blah
Clients:
/...
etc etc
/...

Latency min/avg/max:
Received
Sent
Outstanding
Zxid
Mode
Node count
Connection closed

But instead we see:

stat
Zookeeper version blah blah
Clients:
/..
....
...
/.. Connection closed

I'm at a bit of a loss for how to debug this, whether it is a ZK
command problem, a machine config problem, or what. Anyone have any
ideas?

Thanks,
C

Reply via email to