Redis service hangs with systemctl and Type=notify

Question

I have followed this answer: https://serverfault.com/a/893075/210494. I have CentOS 7.8.

redis.service

[Service]
Type=notify
ExecStart=/opt/redis/bin/redis-server /opt/redis/conf/redis-master.conf
TimeoutStartSec=60
TimeoutStopSec=60
TimeOutSec=90
RestartSec=5s
Restart=on-success

redis-master.conf

daemonize no
supervised systemd

When I run this service, the Redis process starts but the systemctl hangs and I have to press Ctrl-C to get back to shell.

This is in the logs:

69486:C 28 Aug 2020 17:31:14.545 # Redis version=6.0.6, bits=64, commit=00000000, modified=0, pid=69486, just started
69486:C 28 Aug 2020 17:31:14.545 # Configuration loaded
69486:C 28 Aug 2020 17:31:14.545 # WARNING supervised by systemd - you MUST set appropriate values for TimeoutStartSec and TimeoutStopSec in your service unit.
69486:M 28 Aug 2020 17:31:14.546 * Increased maximum number of open files to 4096 (it was originally set to 1024).
69486:M 28 Aug 2020 17:31:14.547 * Running mode=standalone, port=6380.
69486:M 28 Aug 2020 17:31:14.547 # Server initialized

If I comment out Type=notify, then systemctl finishes immediatelly but this is in the logs.

68629:C 28 Aug 2020 17:29:58.933 # Redis version=6.0.6, bits=64, commit=00000000, modified=0, pid=68629, just started
68629:C 28 Aug 2020 17:29:58.933 # Configuration loaded
68629:C 28 Aug 2020 17:29:58.933 # WARNING supervised by systemd - you MUST set appropriate values for TimeoutStartSec and TimeoutStopSec in your service unit.
68629:C 28 Aug 2020 17:29:58.933 # systemd supervision requested, but NOTIFY_SOCKET not found
68629:M 28 Aug 2020 17:29:58.933 * Increased maximum number of open files to 4096 (it was originally set to 1024).
68629:M 28 Aug 2020 17:29:58.934 * Running mode=standalone, port=6380.
68629:M 28 Aug 2020 17:29:58.934 # Server initialized

Where is the problem? I found DigitalOcean tutorial where they do not have Type=notify and I found Redis issues when they successfully recommended Type=notify.

Update

Job for redis.service failed because a timeout was exceeded. See "systemctl status redis.service" and "journalctl -xe" for details.
[root@DB dev]# systemctl status redis
● redis.service - Redis
   Loaded: loaded (/etc/systemd/system/redis.service; enabled; vendor preset: disabled)
   Active: failed (Result: timeout) since Sun 2020-08-30 12:06:58 CEST; 46min ago
  Process: 11134 ExecStart=/opt/app/redis/default/bin/redis-server /opt/app/redis/default/conf/redis-master.conf (code=exited, status=0/SUCCESS)
 Main PID: 11134 (code=exited, status=0/SUCCESS)
Aug 30 12:05:57 DB systemd[1]: Starting Redis...
Aug 30 12:06:58 DB systemd[1]: redis.service start operation timed out. Terminating.
Aug 30 12:06:58 DB systemd[1]: Failed to start Redis.
Aug 30 12:06:58 DB systemd[1]: Unit redis.service entered failed state.
Aug 30 12:06:58 DB systemd[1]: redis.service failed.

Well, that isn't making the slightest bit of sense. Check the journal and `dmesg` to see if there's anything else unusual going on. — Michael Hampton, Aug 30 '20 at 16:19
It would have been helpful to mention that you compiled redis yourself. Nobody would generally expect anyone to do that for a production system. — Michael Hampton, Aug 31 '20 at 13:03
What? Redis does not provide binaries, does it? https://redis.io/download — Leos Literak, Aug 31 '20 at 13:25
If you used a binary then it would have been helpful to mention that too. This also is something that would be unexpected. — Michael Hampton, Aug 31 '20 at 13:26
Though, now that I think about it more, redis official binaries _not_ compiled with systemd support, and not throwing an error when you try to use systemd, is a really bad thing. I suggest filing a bug report. — Michael Hampton, Aug 31 '20 at 13:33
They have incorporated my PR to update the documentation. But throwing an error would be good habbit. I can report it too. — Leos Literak, Aug 31 '20 at 13:36

score 9 · Answer 1 · answered Aug 31 '20 at 09:26

9

Ok, the problem was that systemd support is not compiled by default. You need to compile it explicitelly. I will write a PR to update README.md in sources because this is not explained.

This commit is responsible for this behaviour. I found it in this issue: https://github.com/redis/redis/issues/7217

yum install systemd-devel
make BUILD_WITH_SYSTEMD=yes USE_SYSTEMD=yes

answered Aug 31 '20 at 09:26

Leos Literak

255
2
12

1

Thanks for this. Been debugging this for two hours! – Andy Oct 16 '20 at 13:50

Redis service hangs with systemctl and Type=notify

1 Answers1