Question: Galaxy crash messed up NGINX AND workflow editing tool ?
gravatar for christophe.habib
2.8 years ago by
christophe.habib340 wrote:

Hi everyone.

I have a big issue on a local instance of Galaxy for diagnosis purpose. Here is what happened :

I installed this instance last october and I had many issues du to the quality of the network and the proxy of the hospital where I work. Finally I managed to make things work properly, but I didn't installed/configured any scheduler to manage the jobs.

Yesterday, my colleagues told me they tried to launch the analysis workflow (22 steps, 2 fastq and 1 bed file as inputs) for several patients (like 10, we have a small server) and the instance was no longer accessible. Indeed : blank page when we tried to go to this instance. I went on the IRC channel to find some help.

Thanks to Dannon, we managed to find out that the issue was related to Nginx : all the requests were permission denied :

GET [HTTP/1.1 403 Forbidden 1ms]

My colleagues really needed to have access again to that instance to perform diagnosis, so I managed to get the access to the instance by changing the first line of the nginx.conf from user www-data to user galaxy.

So here is the first thing : I would like to return to the original www-data within nginx because I am afraid that a similar crash might change rights to galaxy itself. Do you have any idea how to give back his rights to www-data ?

The second this is, I thought that everything went back to the normal. BUT when I try to edit an existing workflow I have a message error server. In the paster.log I have this message :

And In galaxy itself we have this error :

Server error
{"upgrade_messages": {}, "steps": {"0": {"tool_id": null, "data_inputs": [], "data_outputs": [{"extensions": ["input"], "name": "output"}], "form_html": "

Several important things regarding this matter :

  • I can export this workflow and the files content is right (and Guerler could import it on his instance)
  • if I import it back it shows the same error.
  • If I create a NEW workflow everything works properly

I really think this is related to the first problem with Nginx. I mean, this workflow was created when the whole galaxy thing was accessible thanks to Nginx which has www-data as worker, not galaxy. Plus the whole galaxy repertory owners were galaxy:www-data at that time. I tried to modify owners to galaxy:galaxy but it didn't fix the problem.

So I think the most important problem is the first one, and the second one will be solve by itself. I would really need your help to fix this. It is such a mess to set up a new instance because of the hospital proxy that I don't want to start from scratch again.

Thank you for your help !

workflow crash galaxy nginx • 1.2k views
ADD COMMENTlink modified 2.8 years ago • written 2.8 years ago by christophe.habib340

I just noticed that when I try to edit a workflow I have this error in the nginx log file:

So my guess may be right ... but I still don't know how to fix nginx !

ADD REPLYlink modified 2.8 years ago • written 2.8 years ago by christophe.habib340

I found another "issue". Windows user can't download file anymore probably because galaxy user is the worker of nginx instead of www-data.

Well I really need to fix this ...

ADD REPLYlink written 2.8 years ago by christophe.habib340
gravatar for christophe.habib
2.8 years ago by
christophe.habib340 wrote:

I did this command

chmod -R g+rwx /home/galaxy/diag_19102015/

And the workflow are editable again. But I dont really like it.

As well, to solve the download issue, i think will have to tive to galaxy all the rights that belongs to www-data to begin with. And i don't like it ... :(

ADD COMMENTlink written 2.8 years ago by christophe.habib340
gravatar for Nate Coraor
2.8 years ago by
Nate Coraor3.2k
United States
Nate Coraor3.2k wrote:

For the record, the permissions problems have been fixed. They were cased by the Galaxy user's home directory (a parent to the Galaxy code directory) being set drwx------ (0700).

ADD COMMENTlink written 2.8 years ago by Nate Coraor3.2k
Yes thank you for your help. I didn't reported it yet because I want to see if I can reproduce the issue. And I didn't have the time to check if the download issue as fixed. I will report all my conclusions next week.
ADD REPLYlink written 2.8 years ago by christophe.habib340

Hi Nate,

I thought that everything worked properly, unfortunately this morning I noticed that the permissions on galaxy went back to drwx------ (0700) when I asked to my colleague to connect on  this instance from his windows XP computer and when he refreshed the welcome page.

I have no explanations about that. I asked my colleague not to use this instance, and I wasn't there the past few days. So nothing from users should have lead to this change (no overload, no misuse, etc.).

How these rights could changed on their own ? This behaviour is a big issue for me...

Once again I do not see where I should start to solve this problem. I checked the /var/log/auth.log and nobody logged in with root to do this kind of bad joke (and I am supposed to be the only one to connect as root).

I hope you will have an idea about that.

Thank you !

ADD REPLYlink modified 2.8 years ago • written 2.8 years ago by christophe.habib340

What are the correct permissions settings for the directories?

ADD REPLYlink written 2.2 years ago by steve30
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 165 users visited in the last hour