I'm archiving wal files from a postgres server via rsync, most of the time the archiving works fine and is quick, the speedtest for the connection is here: (this goes via the internet)
[ ID] Interval Transfer Bandwidth Retr Cwnd
[ 4] 0.00-1.00 sec 9.30 MBytes 78.0 Mbits/sec 0 395 KBytes
[ 4] 1.00-2.00 sec 66.3 MBytes 556 Mbits/sec 14 1.05 MBytes
[ 4] 2.00-3.00 sec 75.0 MBytes 629 Mbits/sec 0 1.16 MBytes
[ 4] 3.00-4.00 sec 81.2 MBytes 682 Mbits/sec 0 1.24 MBytes
[ 4] 4.00-5.00 sec 86.2 MBytes 724 Mbits/sec 0 1.30 MBytes
[ 4] 5.00-6.00 sec 88.8 MBytes 744 Mbits/sec 0 1.34 MBytes
[ 4] 6.00-7.00 sec 91.2 MBytes 765 Mbits/sec 0 1.37 MBytes
[ 4] 7.00-8.00 sec 92.5 MBytes 776 Mbits/sec 0 1.38 MBytes
[ 4] 8.00-9.00 sec 93.8 MBytes 786 Mbits/sec 0 1.39 MBytes
[ 4] 9.00-10.00 sec 63.8 MBytes 535 Mbits/sec 22 535 KBytes
So, more than enough bandwidth available.
But on some WAL files it just crawls to a slow and it takes 30 - 50 seconds until the 16 MB file is transferred, and I'm at a loss where to debug / look for the issue.
The rsync command looks like this:
rsync -p --chmod=Fg+r,Fo+r --timeout 10 -e /usr/bin/ssh -i /var/lib/pgsql/.ssh/id_rsa -a pg_wal/000000080000A5500000005D barman@barman_host/data/database/pg/incoming/000000080000A5500000005D
I looked at rsync via strace on the receiving side, and it seems that just the packets from the sending side were not arriving fast enough. I tried to cat the file via ssh and output it on my console, that was finished before the rsync transfer. I tried to cat it to /dev/null, that was instant.
So I assume the source drive is fast enough.
I transferred a bulk of WAL files (60GB) via a single rsync command, that was also quick and averaged 65 MB/s, so that would tell me that everything works fine, still some files are slow.
What else can I look at? How can I figure out if the issue is sending side, internet speed, receiving side, are there some special logs I can activate on rsync? Can I check the timings of the syscalls via strace?
ls -l 000000080000A578000000E8
-rw-------. 1 postgres postgres 16777216 Jul 19 07:32 000000080000A578000000E8
bash-4.2$ du -sh 000000080000A578000000E8
11M 000000080000A578000000E8
bash-4.2$ du -sh 000000080000A578000000E8 --apparent-size
16M 000000080000A578000000E8
WAL drive is a ZFS with compression active, therefore the difference.
Also for completion, all zfs properties:
storage/database type filesystem -
storage/database creation Thu Apr 19 12:22 2018 -
storage/database used 1.33T -
storage/database available 369G -
storage/database referenced 1.33T -
storage/database compressratio 2.13x -
storage/database mounted yes -
storage/database quota none default
storage/database reservation none default
storage/database recordsize 16K inherited from storage
storage/database mountpoint /data/
local
storage/database sharenfs off default
storage/database checksum on default
storage/database compression lz4 inherited from storage
storage/database atime off inherited from storage
storage/database devices on default
storage/database exec on default
storage/database setuid on default
storage/database readonly off default
storage/database zoned off default
storage/database snapdir hidden default
storage/database aclinherit restricted default
storage/database createtxg 1159021 -
storage/database canmount on default
storage/database xattr sa inherited from storage
storage/database copies 1 default
storage/database version 5 -
storage/database utf8only off -
storage/database normalization none -
storage/database casesensitivity sensitive -
storage/database vscan off default
storage/database nbmand off default
storage/database sharesmb off default
storage/database refquota none default
storage/database refreservation none default
storage/database guid 8214081110063784152 -
storage/database primarycache all default
storage/database secondarycache all default
storage/database usedbysnapshots 0B -
storage/database usedbydataset 1.33T -
storage/database usedbychildren 0B -
storage/database usedbyrefreservation 0B -
storage/database logbias throughput inherited from storage
storage/database dedup off default
storage/database mlslabel none default
storage/database sync disabled local
storage/database dnodesize legacy default
storage/database refcompressratio 2.13x -
storage/database written 1.33T -
storage/database logicalused 2.82T -
storage/database logicalreferenced 2.82T -
storage/database volmode default default
storage/database filesystem_limit none default
storage/database snapshot_limit none default
storage/database filesystem_count none default
storage/database snapshot_count none default
storage/database snapdev hidden default
storage/database acltype off default
storage/database context none default
storage/database fscontext none default
storage/database defcontext none default
storage/database rootcontext none default
storage/database relatime off default
storage/database redundant_metadata all default
storage/database overlay off default
But nothing was changed in the past days on the ZFS drive - and this whole issue just started on Friday (17th July).
Also, if I copy paste the command, and run it again, it finishes instantly - the still running command will continue to hang.
With ls -lah you can follow how the temporary file gets bigger and bigger (around 150 KB/s)
Thanks for anyone taking time to read this!
Edit: I added timing recording to the wal archive process, here's the result:
000000080000A57C00000034 1
000000080000A57C00000035 0
000000080000A57C00000036 0
000000080000A57C00000037 1
000000080000A57C00000038 1
000000080000A57C00000039 119
000000080000A57C0000003A 2
000000080000A57C0000003B 1
000000080000A57C0000003C 127
000000080000A57C0000003D 2
000000080000A57C0000003E 1
000000080000A57C0000003F 1
000000080000A57C00000040 1
000000080000A57C00000041 1
000000080000A57C00000042 1
000000080000A57C00000043 1
000000080000A57C00000044 1
000000080000A57C00000045 1
000000080000A57C00000046 1
000000080000A57C00000047 1
000000080000A57C00000048 1
000000080000A57C00000049 105
000000080000A57C0000004A 2
000000080000A57C0000004B 2
000000080000A57C0000004C 1
000000080000A57C0000004D 1
000000080000A57C0000004E 118
000000080000A57C0000004F 2
000000080000A57C00000050 1
000000080000A57C00000051 120
000000080000A57C00000052 2
000000080000A57C00000053 1
The number on the right is seconds it took to execute the Rsync command for the specified file.
Edit 2:
I have recreated the issue with 2 ram drives on each side. I extracted the ports used and have found they are all even (might be a hint)
I have switched between internet connections on my side (target) and the problem went away. Based on discussions it seems to be a network issue over a certain path (maybe due to Load Balancing)
I will update with the final resolution.
Edit 3:
Our provider is Hetzner, they had a fault in one of their DECIX modules (https://www.hetzner-status.de/#16045). After the deactivation the issue went away.
Since the fault was network related as initially thought, I would recommend anyone to follow the same steps.
Keeping this question open makes no sense, since it's an intermittent issue, but maybe someone in the future might have the same.