Mattermost System Console Doesn't Show Up in Mattermost GUI

Troubleshooting
1

Issue Summary
In Omnibus GitLab docker, I don’t have the System Console available

Steps to reproduce

No Console shows up in the correct place

Expected behavior

Shows up in the mattermost chat window somewhere I presume

Observed behavior

I have seen it turn up randomly once in a non administrator account, then it did disappear. I have tried ensuring roles are available via CLI, however I am hit by both the i18n error, thus:

mattermost --config=/var/opt/gitlab/mattermost/config.json version
panic: Unable to find i18n directory (and many more lines of errors afterward),

And then the following database connection error (my assumption it’s a database connection error, but it looks like one to me):

E.g. if I run from /opt/gitlab/embedded/service/mattermost to work around that, I get:

{“level”:“info”,“ts”:1586416609.6965177,“caller”:“utils/i18n.go:83”,“msg”:“Loaded system translations”,“for locale”:“en”,“from locale”:"/opt/gitlab/embedded/service/mattermost/i18n/en.json"}

{“level”:“info”,“ts”:1586416609.696926,“caller”:“app/server_app_adapters.go:58”,“msg”:“Server is initializing…”}

{“level”:“info”,“ts”:1586416609.7034106,“caller”:“sqlstore/supplier.go:221”,“msg”:“Pinging SQL”,“database”:“master”}

{“level”:“error”,“ts”:1586416609.7047381,“caller”:“sqlstore/supplier.go:233”,“msg”:“Failed to ping DB”,“error”:“dial tcp 127.0.0.1:3306: connect: connection refused”,“retrying in seconds”:10}

Clearly it can’t see my database. However I haven’t been able to figure out what is calling this and why it can’t see my database, when it’s an omnibus edition and the rest of mattermost is working.

Can anyone shed any light on where the correct setting for this database might be?

I did see this line in the config file, perhaps it’s this?

mattermost[‘sql_data_source’] = "mmuser:mostest@tcp(127.0.0.1:3306)/mattermost_test>

I can’t manage much without the system console! :smiley:

Many thanks.

Marshalleq.

2

@paulrothrock Would the support team know about this?

1 Like
3

This is what I’ve found:

If I enable mattermost / gitlab SSO as per instructions, the system console disappears. If I don’t, it remains visible.

I’ve found other annoying things also, making me wonder if this system is really as polished as it needs to be, such as:

Safari caching of the site is very bad - i.e it remembers previous user details within the next user (needs cache to be cleared every time), otherwise you end up calling a user by another users name when you edit them and other even more alarming possibilities.

The gitlab.rb file isn’t always reflected in system console, e.g it will say http in one and https in the other (yes I’ve run the gitlab-ctl reconfigure, cleared the cache and tried new / different browsers)

All of the console based command line tools I’ve tried don’t work except the reconfigure command.

e.g. gitlab-rails console production produces a long line of errors, starting with: Traceback (most recent call last):
8: from bin/rails:4:in `’

Perhaps I’m just not doing this right, but command line tools, a thing that should and ultimately must be simple is not simple. Errors on standard command line tools should not be happening, they should be producing sane errors at very least, not the insane long ones that are more like compiling errors.

I’ve done many reinstalls of this official omnibus docker edition. Perhaps it’s just the ombnibus edition I don’t know, but this should be much better.

One thing that’s interesting is unlike how the docs say to do it, if I disable Mattermost / Gitlab SSO, it actually enables it automatically by default. After reading all the instructions, I really wasn’t expecting that. Perhaps they should be updated to reflect this - perhaps I missed it, I don’t know. It would have saved me a lot of time if I had seen it in t he beginning. Perhaps it’s something new that hasn’t been added to the docs yet. But this makes the system console and SSO work. Go figure. So that one seems solved.

There’s no test email functionality seemingly which is frustrating. The console method does not work because well, the console error issues I mentioned above.

Sorry if this seems overly negative, it has been a frustrating few days. I’ve now got to decide whether to recommend this to the organisation or not and a big worry is even if we accept this as running now, will it stay that way and will it be a major pain to maintain, I’m suspecting it will be quite problematic, but would be interested to know what others think before making that decision.

Also note, I’m not a dev, I’m infrastructure.

Marshalleq