Thread: Is PostgreSQL supported on RHEL6?
Hello, I have some software products which support RHEL5 for x86 and x86_64. Each of them uses PostgreSQL 8.3.12 as a data repository. They all embed the same PostgreSQL binaries. Now I'm trying to support both RHEL5 and RHEL6 with minimal effort (but with safety). If possible, I want to continue to use PostgreSQL 8.3.12 built on RHEL5 for a while. Then, I'd like to ask some questions: Q1: Is it safe to use PostgreSQL 8.3.12 on RHEL6? If it is not safe, what kind of problems might happen? I could build 8.3.12 successfully with 167 compilation warnings that report "variable not used" and "uninitialized variable is used" etc. Even if I could run PostgreSQL, I'm not sure that it is safe. I wonder if running the regression tests reveals problems. I searched the PostgreSQL mailing lists with "RHEL6" and found the discussion regarding wal_sync_method and O_DSYNC/O_SYNC. The following fix in 8.3.13 makes me wonder if I should update with 8.3.14 which is the latest version of 8.3 series. Is it safe to use 8.3.12 on RHEL6 by setting wal_sync_method to fdatasync? ---------------------------------------- 8.3.13 release note http://www.postgresql.org/docs/8.3/static/release-8-3-13.html ... Force the default wal_sync_method to be fdatasync on Linux (Tom Lane, Marti Raudsepp) The default on Linux has actually been fdatasync for many years, but recent kernel changes caused PostgreSQL to choose open_datasync instead. This choice did not result in any performance improvement, and caused outright failures on certain filesystems, notably ext4 with the data=journal mount option. ---------------------------------------- Q2: If 8.3.12 is not safe on RHEL6, is 8.3.14 safe? Do I need to use 9.0.3 on RHEL6? I want to avoid upgrading to a newer major version (9.0) because my software do not need new features in 9.0 yet. Q3: Doesn't PostgreSQL's performance degrade on RHEL6? As stated above, by searching the PostgreSQL mailing lists and other web sites, I knew that O_SYNC was implemented in Linux kernel and fsync() got slower (on ext4 than on ext3?). Do these mean that running PostgreSQL on RHEL6 is not appropriate yet? Regards MauMau
"MauMau" <maumau307@gmail.com> writes:
> Now I'm trying to support both RHEL5 and RHEL6 with minimal effort (but with
> safety). If possible, I want to continue to use PostgreSQL 8.3.12 built on
> RHEL5 for a while. Then, I'd like to ask some questions:
I'd recommend rebuilding the executables on RHEL6 if possible, but
otherwise this should be no problem.
> I could build 8.3.12 successfully with 167 compilation warnings that report
> "variable not used" and "uninitialized variable is used" etc. Even if I
> could run PostgreSQL, I'm not sure that it is safe.
You can reasonably assume those are cosmetic. Newer compilers tend to
be pickier about that sort of thing than older ones, so we fix those
sorts of warnings when we see them. If any of them had represented
actual bugs, we'd have back-patched the fixes into 8.3.x.
> I searched the PostgreSQL mailing lists with "RHEL6" and found the
> discussion regarding wal_sync_method and O_DSYNC/O_SYNC. The following fix
> in 8.3.13 makes me wonder if I should update with 8.3.14 which is the latest
> version of 8.3 series.
That would be a good idea in any case.
regards, tom lane
Hello, Thank you for your reply. I've been largely relieved. I understood the anser was as follows: [A1] Yes, it is safe to use PostgreSQL 8.3.12 on RHEL6. It is recommended to rebuild it on RHEL6, however, it should be no problem to use it without rebuilding it. [A2] N/A because the answer to Q1 is yes (safe). Applying the newest update is always recommended. However, the newest update is not a must for RHEL6. Tom, all, I'm sorry to ask again. Do I need to set wal_sync_method to fdatasync in postgresql.conf if I use 8.3.12 on RHEL6? From the release note, I got the impression that even open_datasync is not a problem unless the ext4 file system is mounted with data=journal. How about Q3? Regards Maumau "Tom Lane" <tgl@sss.pgh.pa.us> wrote in message news:12396.1300373556@sss.pgh.pa.us... > "MauMau" <maumau307@gmail.com> writes: >> Now I'm trying to support both RHEL5 and RHEL6 with minimal effort (but >> with >> safety). If possible, I want to continue to use PostgreSQL 8.3.12 built >> on >> RHEL5 for a while. Then, I'd like to ask some questions: > > I'd recommend rebuilding the executables on RHEL6 if possible, but > otherwise this should be no problem. > >> I could build 8.3.12 successfully with 167 compilation warnings that >> report >> "variable not used" and "uninitialized variable is used" etc. Even if I >> could run PostgreSQL, I'm not sure that it is safe. > > You can reasonably assume those are cosmetic. Newer compilers tend to > be pickier about that sort of thing than older ones, so we fix those > sorts of warnings when we see them. If any of them had represented > actual bugs, we'd have back-patched the fixes into 8.3.x. > >> I searched the PostgreSQL mailing lists with "RHEL6" and found the >> discussion regarding wal_sync_method and O_DSYNC/O_SYNC. The following >> fix >> in 8.3.13 makes me wonder if I should update with 8.3.14 which is the >> latest >> version of 8.3 series. > > That would be a good idea in any case. > > regards, tom lane >