It's not a bad idea, but there's a lot that comes to mind you will want to be careful with. I have ran non-foreman apps on Galera for years with great success and quite serious scale (the biggest is currently running 1.4M QPS across a 6 node cluster). A random list:
* Make sure you have sufficient performance capacity across the cluster. When a node can't keep up with the cluster it goes into a state called flow control, which essentially stops all writes cluster wide until the slow node has caught up. I don't need to explain you will want to tune and make sure flow control _never_ gets triggered. * the real power of Galera comes when you can loadbalance reads across the cluster, but this means the app needs to support intelligent multiple connection strings. It's a long story, but I'll try to be short: ** In Galera you can in theory send any connection to any node and it will have a consistent state (semi-sync replication). In reality, a node can be behind a few contracts and this can lead to reads happening directly after a write not finding what they just wrote. The other problem is hotspots. If many places in the application write to the same database page but those connections go to different nodes in the cluster you can get (essentially) deadlock errors. ** to counter the above symptoms, most custom applications use something like a haProxy load balancer in tcp mode that has two frontends: 1 for writes and reads that need to happen directly after writes, the second for all other selects. However, I'm not aware of the foreman supporting such a thing at present. That means you have two options depending on how much you like to live on the edge: 1) you send all foreman traffic straight up to a single node in the cluster, thereby losing scalability for the foreman app. 2) you send all connections load balanced to all galera nodes (using haproxy if you want) and you cross your fingers that it will all work out. So, all of the above kept in mind, I'd totally give it a go (Galera beats Async replication on many fronts). What would I do to be on the safe side? 1) Set up a galera cluster and make the whole cluster a slave of your existing mysql master. Let that run for a few weeks and monitor closely so you know it can run at least your write load properly 2) Presuming you have a test environment set up another cluster there and take some access logs from production and replay them with a load testing tool to see how that holds up. Once happy do the same on your production galera cluster. 3) Set up haproxy as mentioned above with two frontends, but send all traffic to the frontend that goes to a single node in the cluster 4) Start monitoring where your heaviest/most frequent selects are coming from and hack foreman to send those to your secondary haproxy frontend which loadbalances round robin. 5) submit PR and let all of us profit :) I gave a presentation on galera with some more info here: http://www.slideshare.net/olindata/1-m-qps-on-mysql-galera-cluster Hope this helps! Walter On Sunday, April 2, 2017 at 11:54:45 PM UTC+3, Matt wrote: > > I would like to know as well as I'm still running PgSQL locally and want > to move that to one of my MySQL clusters running Percona. > > As far as I know it should not be any problem, have you tested it already ? > > Thanks, > > Matt > > Op zondag 2 april 2017 15:40:30 UTC+2 schreef Lang, Jason: >> >> Hey All, >> >> >> >> Is anyone running Foreman with Galera? I’ve had my lab environment >> (1.13.3) running on a 3 node MariaDB 10.1 Galera cluster for the past few >> weeks and haven’t noticed any issues thus far. I haven’t seen anything in >> the way of “official/listed” support, and reading into Galera, there >> “should” be no issues other than if Foreman didn’t support innodb, or used >> tables without keys. I’ve also googled a bit and found a few threads that >> seemed to indicate “it should work” and “let us know if it doesn’t” but >> they are older posts at this point. >> >> >> >> That begs my question, is anyone using it? Long-term? Any issues, >> gotchas? Most importantly, is there any reason I shouldn’t use it in my >> production environment? Our foreman environment currently has ~8600 servers >> reporting in which translates to quite a few DB calls during normal >> operation. We’ve been using mysql’s built in replication until now – but >> are looking to upgrade to using Galera if possible for better seamless >> ha/failover between our 3 datacenters. >> >> >> >> Finally, if it’s a bad idea, I’m definitely open to suggestions/advice on >> an HA/load balanced database solution that might be better suited for >> Foreman? >> >> >> >> Thanks for any and all feedback, advice, and suggestions >> >> >> >> ~Jason Lang >> >> The information contained in this message may be privileged, confidential >> and protected from disclosure. If the reader of this message is not the >> intended recipient, or an employee or agent responsible for delivering this >> message to the intended recipient, you are hereby notified that any >> dissemination, distribution or copying of this communication is strictly >> prohibited. If you have received this communication in error, please notify >> your representative immediately and delete this message from your computer. >> Thank you. >> > -- You received this message because you are subscribed to the Google Groups "Foreman users" group. To unsubscribe from this group and stop receiving emails from it, send an email to foreman-users+unsubscr...@googlegroups.com. To post to this group, send email to foreman-users@googlegroups.com. Visit this group at https://groups.google.com/group/foreman-users. For more options, visit https://groups.google.com/d/optout.
