On Wed, Sep 2, 2020 at 7:54 PM Tobias Eriksson <tobias.eriks...@qvantel.com> wrote:
> Hi > > I am monitoring a 20+ node cluster, and 5 of them has an excessive GC > Count > > Typically the nodes has 40-60,000 GC runs, but a handful has 4,000,000 GC > runs > > And it is not temporary this is the same all the time, see diagram below > > Interesting is that this is NOT the machines with high latency, instead > the latency is fine > > But I am curious to how come I have such crazy runs on a handful of the > machines > > Trying to reason I figured > > 1. These machines become coordinators more than any other > 2. The partitioning is skewed, some often used table have partition > keys that end up on these 5 machines > 3. … > > If the count is since the JVM on the node started, could it just be that these nodes were not restarted recently, but others were?.. -- Alex
