This is happening in 5.3.1 This metric is interesting to know the minimal memory footprint of a core (data structures and caches).
I agree with Shawn that if Solr doesn't support the metric should be remove from the admin, but I insist in the fact that it's useful to plot memory consumption in services like zabbix. -- /Yago Riveiro On 23 Sep 2016, 01:08 +0100, Shawn Heisey <apa...@elyograg.org>, wrote: > On 9/22/2016 4:59 PM, Yago Riveiro wrote: > > The Heap Memory Usage in the UI it's always -1. There is some way to > > get the amount of heap that a core consumes? > > In all the versions that I have looked at, up to 6.0, this number is > either entirely too small or -1. > > Looking into the code, this info comes from the /admin/luke handler, and > that handler gets it from Lucene. The -1 appears to come into play when > the reader object is not the expected type, so I'm guessing that past > changes in Lucene require changes in Solr that have not yet happened. > Even if the code is fixed so the reader object(s) are calculated > correctly, that won't be enough information for a true picture of core > memory usage. > > In order for this number to be accurate, size information from other > places, such as Lucene caches and Solr caches, must also be included. > There might also be memory structures involved that I haven't even > thought of. It is entirely possible that the code to gather all this > information does not yet exist. > > In my opinion, the Heap Memory statistic should be removed until a time > when it can be overhauled so that it is as accurate as possible. Can > you open an issue in Jira? > > Thanks, > Shawn >
