sigar_port consuming 1.5gb of ram and 10% cpu usage


#1

After upgrading to Couchbase 2.1.0 on a Windows Server 2012 machine with 2 memcache buckets and 2 couchbase buckets yesterday at 16.00, we this morning found the sigar_port process consuming alot of RAM and CPU.

The buckets have no traffic at all under this period.

We restarted the services which helped a bit, but sigar_port is again climbing in memory usage. Is this common?


#2

Thank you very much for reporting this.

We found this right after we cut final RC: http://www.couchbase.com/issues/browse/MB-8465 but in our testing we did not observe any significant leak. So we decided that RC is good enough. I’m very sorry for that.

But good news is that you can use sigar_port.exe from 2.0.1 package. You can either extract it yourself, or you can get whatever I just extracted and successfully verified with 2.1.0: https://docs.google.com/file/d/0B6NtGsLhIcf7b2pfWV82OV9HWmc/edit


#3

On 2.1.1 I’m seeing sigar_port.exe using 100% CPU (100% of a whole core on an 8-core machine, so showing up as 13% in Task Manager). Memory usage looks ok though at about 3MB.

I tried taking the node off the cluster and even leaving it uninitialized (i.e. before going through the web wizard to create/join a cluster), I still see high CPU usage.


#4

That’s definitely unexpected. Please submit bug report with output of cbcollect_info to http://www.couchbase.com/issues/secure/Dashboard.jspa


#5

I’ve since tried reinstalling Couchbase but the installation is hanging at “Computing Space Requirements” step. If I manage to get it installed I’ll try to reproduce it.


#6

just created http://www.couchbase.com/issues/browse/MB-9121 with cbcollect_info on another node