> > Why have a cold backup and then switch? > my current set up is: 1. master indexer 2. master slave on a release/commit basis 3. 3 live slave searching nodes in two data different centers
the three live nodes are in front of nginx load balancing and they are mostly hot but not all of them, i found that having all load into one made the performance significantly better, but if one of them goes down theres a likelihood that the other two went with it, they are also part of a mysql gallera cluster and it has a possibility of going down (innodb can be annoying), so the script will go through all three of the live slaves until it has to fall back to the master slave, i know the cold master will work, mostly out of faith, but if i lose four servers all at the same time, i have larger problems to worry about than searching. just adaptation over time, I cant say its the best set up but i can say it operates pretty well, very well speed wise keeping one searcher super hot with two clones ready to jump in if needed On Mon, Jun 8, 2020 at 12:30 PM Walter Underwood <wun...@wunderwood.org> wrote: > I could write a script, too, though I’d do it with straight shell code. > But then I’d have to test it, check it in somewhere, document it for ops, > install it, ... > > Instead, when we switch from monit, I'll start with one of these systemd > configs. > > https://gist.github.com/hammady/3d7b5964c7b0f90997865ebef40bf5e1 < > https://gist.github.com/hammady/3d7b5964c7b0f90997865ebef40bf5e1> > > https://netgen.io/blog/keeping-apache-solr-up-and-running-on-ez-platform-setup > < > https://netgen.io/blog/keeping-apache-solr-up-and-running-on-ez-platform-setup > > > https://issues.apache.org/jira/browse/SOLR-14410 < > https://issues.apache.org/jira/browse/SOLR-14410> > > Why have a cold backup and then switch? Every time I see that config, I > wonder why people don’t have both servers live behind a load balancer. How > do you know the cold server will work? > > wunder > Walter Underwood > wun...@wunderwood.org > http://observer.wunderwood.org/ (my blog) > > > On Jun 8, 2020, at 9:20 AM, Dave <hastings.recurs...@gmail.com> wrote: > > > > A simple Perl script would be able to cover this, I have a cron job Perl > script that does a search with an expected result, if the result isn’t > there it fails over to a backup search server, sends me an email, and I fix > what’s wrong. The backup search server is a direct clone of the live server > and just as strong, no interruption (aside from the five minute window) > > > > If you need a hand with this I’d gladly help, everything I run is Linux > based but it’s a simple curl command and server switch on failure. > > > >> On Jun 8, 2020, at 12:14 PM, Jörn Franke <jornfra...@gmail.com> wrote: > >> > >> Use the solution described by Walter. This allows you to automatically > restart in case of failure and is also cleaner than defining a cronjob. > Otherwise This would be another dependency one needs to keep in mind - > means if there is an issue and someone does not know the system the person > has to look at different places which never is good > >> > >>> Am 04.06.2020 um 18:36 schrieb Ryan W <rya...@gmail.com>: > >>> > >>> Does anyone have a script that checks if solr is running and then > starts it > >>> if it isn't running? Occasionally my solr stops running even if there > has > >>> been no Apache restart. I haven't been able to determine the root > cause, > >>> so the next best thing might be to check every 15 minutes or so if it's > >>> running and run it if it has stopped. > >>> > >>> Thanks. > >
