Hello cluster techies,
We are running a 5 node Percona Xtradb cluster, all runnning in the same rack, connected with a 1 gb lan connection. We are using the cluster for some larger websites that do a lot of read and writes. Everytime we need to duplicate or import a new database, our website get timeouts when the do a select. It looks like all databases are “offline” while doing this import.
Is there something wrong with our cluster, is this normal or did we do something wrong?
Server specs:
12 cores
64GB memory
750 GB hardware raid
Centos Linux
Percona XtraDB 5.6
Google didn’t help me find anything about it, I really hope you can.
Per the manual - [url]Percona XtraDB Cluster
[LIST]
[*]The write throughput of the whole cluster is limited by weakest node. If one node becomes slow, whole cluster is slow. If you have requirements for stable high performance, then it should be supported by corresponding hardware.
[/LIST] So this might explain that when you start the ‘heavy’ SELECT transaction it slows down the rest of the cluster nodes.
One way to check this is to use Jay Janssen’s myq-tools ([url]https://github.com/jayjanssen/myq-tools[/url]) while running the SELECT query. So you can have a better view of what is happening under the hood.
Thank you jrivera. But is it normal selects are not working while doing an import? io cpu and network usage are low on the servers.
Hi,
Selects could be paused if the cluster went into Flow Control condition. This may happen when some transactions are being applied with too much delay on some (or all) of the slave nodes (nodes replicating from writer node(s)).
From your description, I assume you are doing some large imports, so probably huge transactions are involved. Note that this is not good for Galera based clusters, and you should probably split large imports into smaller chunks.