-
Notifications
You must be signed in to change notification settings - Fork 631
Description
Describe the bug
After issuing a service start, the systemd service start job never completes if certain conditions.
Depending on the issue, or how the service is configured (type=notify vs type=forking) it may not respond to service stop either.
To reproduce
Create a single node, configured as active-replica. Set the replicaof to a non-existent server to replicate a broken setup.
Start service: systemctl start keydb-server
Observe: systemctl list-jobs
You should notice a job is "stuck" in the start running state.
Check the journal. You should see expected errors:
MASTER <-> REPLICA sync started
Connecting to MASTER .....
Error condition on socket for SYNC: Connection refused
Try to stop the service: systemctl stop keydb-server
Systemd job will hang "forever".
Try to send a SIGTERM, keydb-server will ignore it.
Only way to reliably stop keydb-server is a SIGKILL.
Expected behavior
service should accept and respond to a SIGTERM during startup.
Additional information
Running debian buster, keydb package provided by EQ-Alpha