0

I setup 4 disks RAID0 in Azure VM (D11, 2 cores, 14GB ram) and moved pg_xlog to the RAID array and the rest of the pgsql/data on the OS disk which is a single disk. I wonder can I configure effective_io_concurrency=4 since pg_xlog is in 4 disks RAID? However the rest of pgsql/data is in single disk. Would the performance be better to move everything include pg_xlog to the RAID? I saw a lot of suggestion to move pg_xlog and the pgdata into different disk but is it matter if everything is in RIAD0?

Ken
  • 13
  • 2
  • 1
    `effective_io_concurrency` only affects bitmap index scans at the moment (9.5). So I wouldn't worry. You do understand that *any fault* in *any disk* will cause *total loss of all your data* in your current configuration, right? Including any of the disks `pg_xlog` is on? Only do this if you can afford to replace your data by reconstructing it from another source, or if you are doing WAL archiving / streaming replication for replication-based redundancy. In which case you might as well disable crash safety too. – Craig Ringer Aug 23 '15 at 10:22
  • I suppose it would be safe to use RAID0 in Azure since it will be replicated by Azure for 3 copies? it is mentioned here. http://blogs.msdn.com/b/igorpag/archive/2014/10/23/azure-storage-secrets-and-linux-i-o-optimizations.aspx – Ken Aug 23 '15 at 15:23
  • 1
    OK, so you're using RAID 0 over network block storage. That's a bit different to doing it on local storage. In that article I note that it strongly recommends throughout that you not use the OS disk for other things if performance matters to you. So yes, you probably should put the main datadir on a separate RAID. – Craig Ringer Aug 24 '15 at 00:20

0 Answers0