Answer the question
In order to leave comments, you need to log in
Is it worth storing transaction files separately from databases?
Comrades! No answer on google! Tell me, is there any point in postgresql to separate the directory of transaction files and the directory of bases on split arrays?
If so, what should be the arrays?
Please note that one base is production, and the other is test based on production. And if it is absolutely specific, then 1s.
Answer the question
In order to leave comments, you need to log in
Do you mean pg_xlog (pg_wal in 10) or pg_clog (pg_xact)? Transaction (metadata) files are second and are not written very intensively. The first one is write ahead log - a write log, which reflects all writing activity of the base and usually synchronous write.
WAL write disk performance is a key factor in commit time. But WAL is written sequentially and due to this, you can keep a pretty good load by placing them exclusively on separate HDDs, not SSDs. Having saved, respectively, a little money on the SSD capacity required for the base.
Well, or if you have a lot of records with normally configured checkpointer and bgwriter, so that there is a need to allocate several separate SSDs for WAL. However, NTFS is mentioned in your tags. They still don’t do normal production at such a normal production, so it’s not your case.
Arrays - RAID10 or RAID1 depending on how many drives there are. For test machines, see for yourself.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question