centstorage loses some perfdata

[centzkus]

Hello,
please can you help me

It looks like the centstorage daemon loses some perfdata. Such perfdata are then missing in both the RRDs and the data_bin table. There are then white spaces on the graphs. It is more often on some services. I can't see any reason why it happens.
I verified that the process-service-perfdata command puts those perfdata into the file. I let centstorage drop perfdata into the drop file. The drop file misses those perfdata too.
Nothing is missing in the nagios_servicechecks table.
There are no errors in centstorage.log

I have centreon 2.1.4 + nagios 3 on debian lenny.

[AkHeNaToN]

Are you sure that centstorage loses perfdata. I think that Nagios have much latency.

[centzkus]

The kern.log is full of lines like this:
nagios kernel: [14918330.801836] centstorage[7372]: segfault at 7f879e6c0958 ip 7f879d7b4ef9 sp 7fffa66b5b88 error 6 in libc-2.7.so[7f879d738000+14a000]

[AkHeNaToN]

Can you try to update "libc" ?

[centzkus]

There were two problematic services. Whenever centstorage proceeded them there was the error and segfault and quite a few other services were lost. When a service was unfortunate to by scheduled near one of the two problematic services its graph was missing. After some time, when such service was shifted a little from the problematic service its graph reappeared.
The two problematic services had the record in centstorage.index_data but had no records in centstorage.metrics. The field metrics in the table under Administration-Options-CentStorage-Manage was empty for the two services. I chose "Empty all Service Data" for the two services and the errors stopped. The problem could start when I enlarged "PRDTool database size" and "Retention Duration for Data in MySQL" and let "Rebuild RRD Database". I did it for all services at the same time.

Now all graphs are almost perfect. But occasionally there is a narrow gap in a graph. It is every time only one isolated row of perfdata which is lost. Since the minimal heartbeats of all RRDs are equal two steps three points are missing in the graph.
It looks like it can hit any service and I can find no rule nor reason.