nrpe check for graylog-api fails with 307 redirect

[Michał Ajduk]

Graylog is running correctly but the API check charm reports false positive.

graylog version: 3.0.1

The nrpe check for graylog API checks url at 127.0.0.1:9000/api:
root@graylog-3:~# /usr/lib/nagios/plugins/check_http -I 127.0.0.1 -p 9000 -u /api -s cluster_id
HTTP CRITICAL: HTTP/1.1 307 Temporary Redirect - string 'cluster_id' not found on 'http://127.0.0.1:9000/api' - 230 bytes in 0.004 second response time |time=0.004246s;;;0.000000;10.000000 size=230B;;;0

This fails with temporary redirect:
< HTTP/1.1 307 Temporary Redirect
< Location: http://127.0.0.1:9000/

Graylog is listening so it is graylog process sending the redirect:
root@graylog-3:~# netstat -lntp | grep 9000
tcp6 0 0 :::9000 :::* LISTEN 3873/java

The check does not seem to use the credentials created in Graylog.

[Diko Parvanov (dparv)]

managed to reproduce; issues is that query must be issued to http:/127.0.0.1:9001/api and not to http:/127.0.0.1:9000/api

[Paul Goins (vultaire)]

Diko, are you sure you were running with channel=3/stable? I don't think port 9001 will even have a listener if running Graylog 3. The REST API was supposed to be served on the same port as the web interface in GL3, which should be port 9000.

[Diko Parvanov (dparv)]

Used default config that comes with the charm which is channel=2/stable

Using 3/stable the issue is still there (redirect + and cat be fixed by changing the NRPE check as follows:

/usr/lib/nagios/plugins/check_http -I 127.0.0.1 -p 9000 -u /api/cluster -s cluster_id -a admin:ADMIN_PASSWORD

So I am guessing it would be either better to check for channel and set different NRPE checks or change the default to 3/stable and comment that 2/stable is no longer supported by the charm.

[Xav Paice (xavpaice)]

I've been unable to reproduce this with the current release and both Graylog 2 and 3. Noting that the bug report is for graylog version: 3.0.1, which listens on port 9000 only.

This is the result from my test machine:

root@juju-3b6299-9:/etc/nagios/nrpe.d# /usr/lib/nagios/plugins/check_http -k Accept:\ application/json -I 127.0.0.1 -p 9000 -u /api -s cluster_id
HTTP OK: HTTP/1.1 200 OK - 447 bytes in 0.007 second response time |time=0.007178s;;;0.000000;10.000000 size=447B;;;0

Could you maybe share the bundle and charm version that was used to generate this issue?

[Michał Ajduk (majduk)]

I have just reproduced this issue.

Charmed version 4.

App Version Status Scale Charm Store Rev OS Notes
graylog 3.0.1 active 1 graylog jujucharms 4 ubuntu

Version:
  graylog:
    charm: cs:~llama-charmers-next/graylog-4

This is the default version included in the templates.

[Michał Ajduk (majduk)]

The provlem was intermittent with this release so I believe that was due to a charm update.

[Nikita Koltsov (nkoltsov)]

I faced exactly the same issue when set graylog to use https charm version is 51 and the issue is still there.

https://git.launchpad.net/charm-graylog/tree/src/reactive/graylog.py#n1323 127.0.0.1 is hardcoded and when apache with https is enabled graylog automatically redirect to config[web_endpoint_uri] which is causing check to fail.