On 11/23/11 14:37, Pawel Sawicki wrote: > Hi All, > > We are running Firebird-2.5.1 on virtual machine with RHEL6.0. Our > hypervisor is KVM. From time to time we get Firebird unresponsive and we > need to restart it. This situation occurs on heavy loaded system, strace > shows that the problem is related with the fragment of code which contains > invocation of pthread_cond_timedwait sytem call (file > /src/jrd/isc_sync.cpp). It looks like all threads lost completely > synchronization and block each other. We use your standard distribution > package. There is small chance that this problem can be attributed to > different environment used during build, thus we are going to recompile > Firebird from scratch. Hovewer more viable explanation is that we > encountered bug which becomes visible in virtual environment. We have some > experiances with timekeeping in virtual machines which is not very accurate > and could possibly lead to this effect. > If pthread_cond_timedwait sytem is not working correctly can it be simply > replaced by other system call which could be more reliable ? Sooner of all pthread_cond_timedwait() is hard to be replaced with something in such context. Can you first of all provide stack backtraces for a hanged server?
------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity, and more. Splunk takes this data and makes sense of it. IT sense. And common sense. http://p.sf.net/sfu/splunk-novd2d Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
