Mentioned in SAL (#wikimedia-operations) [2017-02-28T17:04:18Z] <gehel> restarting blazegraph on wdqs1001 - T159245

I've bumped the memory so it should be better now. ExitOnOutOfMemoryError is an interesting possibility. @Gehel let's talk about it.

An early report based on GC logs is available. This report analyses only 14h of data, so no hard conclusions yet. Still, a few things to note:

• We use G1, I have not much experience with it, I need to learn!
• Over the period analysed, we rarely go over 8Go of heap after GC. This tends to indicate that 16Go is overkill and probably has negative influence on throughput. More data is needed to validate this hypothesis.
• Over the period analysed, we have an average allocation rate of ~400Mo/sec. This is high! Not crazy high, but still high. I suspect that some of the issues we have seen might be related to increase activity (user traffic, updates, ...) which would increase the allocation rate to something higher than what the GC can cope with. This is pure speculation at this point. This is bad news, as it looks like this might become our bottleneck and there is no easy way to overcome memory throughput issues.