Re: Segments in solr index gets deleted after every restart

[Erick Erickson <er...@gmail.com>]

You simply cannot run Solr (well, actually Lucene) with near 100% disk
utilization. When segments are merged, they are _first_ copied to the
the merge destination and, when successful, the old segments are deleted.

So if segments 1 and 2 are being merged, they are copied to segment 3.
So temporarily you'll have two copies of the segments being merged.

What you're probably seeing is that a merge starts to happen, hits disk
full and on restart the files associated with the failed merge are being
deleted.

In fact, to be safe you need to have at least as much free space on your
disk as your index occupies.


Best,
Erick

On Mon, Aug 29, 2016 at 10:48 PM, Nisha Menon <ni...@gmail.com> wrote:
> I have an issue with respect to increasing space on my Solr servers.
>
> Whenever I see that the allocated space is almost full for a particular
> server, I perform a service solr restart and this clears up some space and
> things work normally from there for some time. Again this builds up and I
> get a space utilization 100% warning.
>
> While debugging this, I found that for every Solr restart, for a couple of
> collections, there are some segments that gets deleted. Which means, for
> one segment, its corresponding .nvm, .fdx, .tvx, .si etc gets deleted.
>
> eg: This is one such segment that was deleted after the restart: _on0.nvm, _
> on0.si, _on0_Lucene50_0.dvm,_on0.fnm,_on0.fdx,_on0.tvx,
> _on0_Lucene50_0.tip, _on0.nvd, _on0_Lucene50_0.tim, _on0.fdt,
> _on0_Lucene50_0.dvd, _on0_Lucene50_0.doc, _on0_Lucene50_0.pos, _on0.tvd
>
> Can anyone please explain what probably would be happening behind the
> scenes for such a behavior? I have Solr Cloud with Solr version being 5.2.1
> and lucene version 5.2.1.
> Thanks in advance.