cromwell server and mysql database configuration questions

myourshawmyourshaw University of ColoradoMember

I'm planning for the installation of a pipeline supported by a new (Slurm-ish) cluster. I have a few questions about the cromwell server and the call caching mysql database.

1) Is it a better practice to have an always-running cromwell server, or to run instances of the server on an adhoc basis?

2) Can multiple instances of the cromwell server use the same mysql server without conflict?

3) Can the mysql server safely be used by cromwell and other applications (assuming adequate bandwidth)?

4) Does the cromwell server have to run on the cluster login node, so it can directly issue srun and other Slurm commands, or can it run on another machine and issue Slurm commands via SSH or another means? [I don't see much about Slurm on the forum; for the purposes of this question Slurm is the same as SGE or LSF]

Issue · Github
by Geraldine_VdAuwera

Issue Number
1819
State
open
Last Updated
Assignee
Array

Answers

  • jgentryjgentry Member, Broadie, Dev ✭✭

    Hi -

    re #1: It really depends. I think there are arguments either way. I tend to picture a server as being better for groups that want to be able to manage/track workflows over time.

    re #2: Yes if you're very careful about how you do it. I wouldn't recommend it without talking through your setup with us first.

    re #3: Yes. I'd recommend keeping them in separate schemas but even then as long as the tables are separate it shouldn't be an issue

    re #4: If you're using Slurm, at least at the moment that's going to imply the so called Config backend. If you can express what you need as a command line it can be done. Certainly being able to just say srun would be easier but you should be able to use ssh to execute the command. Warning: We've never tried doing that and we don't have access to a Slurm cluster

Sign In or Register to comment.