On Thu, Oct 17, 2013 at 5:04 PM, John Chilton chilton@msi.umn.edu wrote:
On Thu, Oct 17, 2013 at 5:49 AM, Peter Cock p.j.a.cock@googlemail.com wrote:
On Thu, Oct 17, 2013 at 11:36 AM, John Chilton chilton@msi.umn.edu wrote:
(And with the job splitting not being enabled by default, I am aware that I am in a relatively small group of Galaxy admins using it.)
I don't think my fix hurts non-task split jobs, but I will now try your fix on the default branch:
I think it will in at least some cases if metadata is getting set externally, I don't see how it is preventing some commands from running together, I could totally be wrong though.
At any rate, I tested the version with my fix on your blast wrappers, with task splitting on and off, submitting to a DRM and using the local job runner, and they all seemed to work. Let me know if your galaxy instance is unconvinced.
Seem fine so far :)
This appears to also address a metadata issue, which if I am lucky may be the fix for this issue?:
http://lists.bx.psu.edu/pipermail/galaxy-dev/2013-October/017031.html
I doubt your luck is so good. That problems looks like some sort of disk caching issue to me (galaxy process and worker node having inconsistent views of the same file system), I doubt this will fix it. Though hopefully I am wrong on both counts :).
Disk caching makes sense as a root cause - I've not had this happen consistently or reproducibly yet so it may well return.
Thanks for reporting the problem and the fix!
-John
No problem, thank you for addressing it so promptly.
Peter