Re: BUG #15636: PostgreSQL 11.1 pg_basebackup backup to a CIFS destination throws fsync error at end of backup

Hi Thomas,

I was able to test the patch without issue and it appears to have worked
see below. Let me know if there is further testing I can do or logging you
would like.

existing bin:

[postgres(at)server]$ /usr/pgsql-11/bin/pg_basebackup -D
/cifs/backups/99testNoPatch -U backupuser -Ft -Z 1 -X fetch -p 5433

pg_basebackup: could not fsync file "/cifs/backups/99testNoPatch": Invalid
argument

patched bin:

[postgres(at)server]$ /home/postgres/patchinstall/bin/pg_basebackup -D
/cifs/backups/99testWithPatch -U backupuser -Ft -Z 1 -X fetch -p 5433
[postgres(at)server]$

I also restored the from the patched backup no issues:

2019-02-15 10:29:12.426 EST, LOG, "ending log output to stderr",,"Future
log output will go to log destination ""csvlog""."
2019-02-15 10:29:12.431 EST, LOG, "database system was interrupted; last
known up at 2019-02-15 10:18:00 EST"
2019-02-15 10:29:12.605 EST, LOG, "starting point-in-time recovery to
2019-02-15 10:29:09-05"
2019-02-15 10:29:12.796 EST, LOG, "restored log file
""000000010000000000000008"" from archive"
2019-02-15 10:29:12.845 EST, LOG, "redo starts at 0/8000028"
2019-02-15 10:29:12.848 EST, LOG, "consistent recovery state reached at
0/80000F8"
2019-02-15 10:29:12.849 EST, LOG, "database system is ready to accept read
only connections"
2019-02-15 10:29:13.013 EST, LOG, "restored log file
""000000010000000000000009"" from archive"
2019-02-15 10:29:13.039 EST, LOG, "redo done at 0/9000140"
2019-02-15 10:29:13.135 EST, LOG, "restored log file
""000000010000000000000009"" from archive"
2019-02-15 10:29:13.188 EST, LOG, "selected new timeline ID: 3"
2019-02-15 10:29:13.228 EST, LOG, "archive recovery complete"
2019-02-15 10:29:13.363 EST, LOG, "database system is ready to accept
connections"

Thanks,

John

On Fri, Feb 15, 2019 at 12:53 AM John Klann <jk7255(at)gmail(dot)com> wrote:

> Great thank you I will let you know how is goes.
>
> John
>
> On Thu, Feb 14, 2019, 23:47 Thomas Munro <thomas(dot)munro(at)enterprisedb(dot)com
> wrote:
>
>> On Fri, Feb 15, 2019 at 5:27 PM John Klann <jk7255(at)gmail(dot)com> wrote:
>> > Happy to test it, I will need the instructions if you can. It's for
>> Rhel 7.5 Maipo latest. (Linux 3.10.0-862.14.4.el7.x86_64 #1 SMP Fri Sep 21
>> 09:07:21 UTC 2018
>> > x86_64 x86_64 x86_64 GNU/Linux)
>>
>> Thanks! Use yum to install gcc, make, bison, yacc, git,
>> readline-devel, zlib-devel. (Hope I'm not forgetting anything
>> here...)
>>
>> Pull down the sources:
>>
>> $ git clone https://github.com/postgres/postgres.git
>>
>> Go into the source directory:
>>
>> $ cd postgres
>>
>> Let's say we want to test with the 11 branch (so you can test against
>> an existing 11.x development environment database you might have
>> installed the usual way with packages, as opposed to the master branch
>> where we are developing 12). Switch to that branch:
>>
>> $ git checkout REL_11_STABLE
>>
>> Create a new branch for this experiment:
>>
>> $ git checkout -b tolerate-einval
>>
>> Apply the patch (adjust path to wherever you put the patch):
>>
>> $ git am 0001-Tolerate-EINVAL-when-calling-fsync-on-a-directory.patch
>>
>> Configure, build, install (the prefix is where it will install the
>> resulting binaries, adjust to taste, -j is how many CPUs you have if
>> you want to build fast):
>>
>> $ mkdir $HOME/install
>> $ ./configure --prefix=$HOME/install
>>
>> $ make -s -j4 && make -s install
>>
>> Optionally, run a self-test to see if everything is sane:
>>
>> $ make -s check
>>
>> Now you should have binaries like $HOME/install/bin/pg_basebackup and
>> you can test if they work on CIFS.
>>
>> --
>> Thomas Munro
>> http://www.enterprisedb.com
>>
>