After upgrading to 1.5 and migrating the index to tsi1 influx won't start completely

1

I’ve upgraded to 1.5 and can’t get past this point. Nothing happens here. When I try to connect to influx CLI, I get this:

Please check your connection settings and ensure 'influxd' is running.```

```ts=2018-03-12T22:56:16.478494Z lvl=info msg="InfluxDB starting" log_id=06o6hWNW000 version=1.5.0 branch=1.5 commit=6ac835404e7e64ea7299a6eebcce1ab1ef15fe3c
ts=2018-03-12T22:56:16.478554Z lvl=info msg="Go runtime" log_id=06o6hWNW000 version=go1.9.2 maxprocs=8
ts=2018-03-12T22:56:16.480736Z lvl=info msg="Using data dir" log_id=06o6hWNW000 service=store path=/var/lib/influxdb/data
ts=2018-03-12T22:56:16.480779Z lvl=info msg="Open store (start)" log_id=06o6hWNW000 service=store trace_id=06o6hWO0000 op_name=tsdb_open op_event=start```
2

I have exactly the same issue. Upgraded from 1.4 to 1.5 with dpkg on a debian system. Nothing in the config is changed, except that I migrated to tsi1. Not matter what I do, there’s simply no tcp socket open.

What can I do?

3

Did you try ulimit -n 256789 and then start influxd in that shell environment? This fixed it for me, however, adding the line LimitNOFILE=256789 into influxdb.service does not seem to have any effect… Wonder why…

So, basically, when you start influx via systemd it dies because of too many open files, but that is not detected by systemd. So you think the service is running, but the port is closed. In reality, influxd never makes it through startup.

1 Like
4

Adding LimitNOFILE to influxdb.service AND running systemctl daemon-reload solved the issue. Thanks alot!