Re: Incorrect fillFactor/memory usage metric
Thanks very much for your thoughts on this. I can confirm that the use of getTotalUsedPages() does indeed allow us to calculate a memory consumption metric that reduces as cache data is purged. I will comment on and close the Jira ticket. This is great news. Just one thought, in case there's any work in this area still under way - perhaps it would be worth adding explicit methods on the metrics interface to get used/free memory as per this calculation? It strikes me that these metrics are absolutely critical for many use cases - but obtaining them is not currently obvious or always reliable. I'm very happy for my test (see jira) to be included in the test suite if that's useful? Regards, Colin. -- Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
Re: Incorrect fillFactor/memory usage metric
Hello Colin, AFAIK fill factor doesn't take into account reuse bucket by design, it's not a bug. To calculate memory consumption you can also use metric totalUsedPages which take into account the reuse bucket (but doesn't take into account free-list fill factor) This metric was introduced in Ignite 2.8 and hasn't released yet. . вт, 14 янв. 2020 г. в 19:44, coliin : > Regarding the below issue, which causes incorrect memory consumption metric > in Ignite 2.7+ and related GG versions: > > https://issues.apache.org/jira/browse/IGNITE-12096 > > As reported in the jira comment, this appears to be caused by failure to > account for pages that are flagged for reuse - the FreeList REUSE_BUCKET is > not counted when determining freeSpace. > > Is anyone able to comment on whether there is a good reason for not doing > so? If there isn't then the issue is simple to fix. > > Regards, > Colin. > > > > -- > Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/ >
Incorrect fillFactor/memory usage metric
Regarding the below issue, which causes incorrect memory consumption metric in Ignite 2.7+ and related GG versions: https://issues.apache.org/jira/browse/IGNITE-12096 As reported in the jira comment, this appears to be caused by failure to account for pages that are flagged for reuse - the FreeList REUSE_BUCKET is not counted when determining freeSpace. Is anyone able to comment on whether there is a good reason for not doing so? If there isn't then the issue is simple to fix. Regards, Colin. -- Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/