Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    text
    copied!<p>AFAIK the 'lockfile' - is only tested for existence. The file exists, searchd is activly serving the index. File doesnt exist, the index is safe to be recreated in place or deleted etc. </p> <p>indexer will refuse to reindex an index, if the lockfile exists.</p> <p>But if you specify --rotate, it will instead create a brand new version of the index (with .new. in the filename), it will not care if an lock file exists because it doesnt touch the exisiting index. </p> <p>and when finished signal to searchd. searchd will then delete the active index, rename the index files, and serve the new versions (in such a way to not interupt serving - and it also keeps the lockfile in place) </p> <hr> <p>So from that the only way to stop indexer (assuming you using --rotate), is to create indexname.new.spl - I think that might work - but never tried. I dont think it will notice if you advisory lock something. </p> <p>A better, but more trickly way to get consistant backups, might be to intercept the signal from indexer to searchd. Have a process that listens to the sighup from indexer, if an backup is in progress pause until it finished, and then send the sighup to searchd. </p> <p>(But it might take some work to make indexer send the sighub to your intercepter, will have to create a config file, with a fake pid file. so its the pid of the interceptor not searchd) </p> <hr> <p>Of course the bigger question, if your indexes can be recreated so easily (you recrating them often anyway), why bother backing them up? If lost then can just be recreated. </p>
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload