RepEX cannot execute (04/23/15 morning fresh checkout). #23
Comments
|
Hi Taisung, I just tried to run 2d but I didn't get this error, it went through normally. In the line above the last line it says "Pilot error: [, , , ]" so I guess maybe it's some issues with radical.pilot rather than repex. By "fresh checkout" do you mean both repex and radical.pilot? Also, the warning in the beginning might suggest something. Good luck! |
|
I did fresh checkout for everything. Taisung |
|
Taisung, what is the version of radical pilot? e.g.: radicalpilot-version in terminal? Error is not very informative, so I don't really understand what went wrong. Do you have problems with 2d example or with temperature exchange as well? Was Pilot actually launched on the machine? or failure occured even before that? If pilot was launched, can you please provide logs? |
|
I just tried to reproduce it but it works now…. I will try to figure out what is going on. Taisung |
|
Just tried it and the same error happened again. The job was not launched on the remote machine (Stampede). Taisung |
|
Taisung, Thanks, Andre. |
|
This is the output of It seems that after some point on 4/13, the radical.pilot has been changed. For those Taisung |
|
The change in names comes from the recent release we did. You seem to indicate that you did a 'fresh checkout' (it says so in one of the earlier messages). Does that also include creating a fresh virtualenv on your local machine? If not, please try that -- it might be that only a part of the stack got installed after the new release. If you tried that already, then I would like to ask you for a log file from your client machine, ie. the output you get when running with Many thanks, Andre. |
|
See log below. Every module was fresh checked-out. The virtualenv was newly created. |
Note that this is not true, but Taisung found them by now I think. |
I'm seeing a similar thing on a different machine, seems that sdist is not created (correctly). |
Yes, indeed, but I can't reproduce this, yet. Do you have an env or shell script for me which triggers that error? The version of python, pip and setuptools would be interesting, too, I guess. |
|
The sdist should not be required in the first place. This points to wrong defaults in the resource configuration file(s). With regards to the absence of the sdist, that only gets created I think when you actively install the dependencies from source, not when they are pulled in. |
|
Ack on the resource config -- but I can't reproduce the problem with the sdists: Any other idea how to reproduce this? In what context were the sdists missing for you? |
Created radical-cybertools/radical.pilot#578 for that.
I actually tried, but failed, can't reproduce it anymore. |
|
Maybe Taisung can ... |
|
Taisung, if you continue to have trouble with the current release and if that stop you from doing work, you may want to revert to the previous version: should give you (with But we would still be interested to know, of course, what caused your troubles, so please let us know if you have any details on how you installed the new radical.pilot release (and dependencies). Best, Andre. |
|
Antons seemed to identity the problem: the problem will occur if I run “python setup.py install” for each module, but will be OK if I use “pip install --upgrade .” instead. You may try if you can reproduce it. Taisung From: Andre Merzky [mailto:notifications@github.com] Taisung, if you continue to have trouble with the current release and if that stop you from doing work, you may want to revert to the previous version: $ virtualenv ve should give you (with RADICAL_VERBOSE, SAGA_VERBOSE and RADICAL_PILOT_VERBOSE all set to DEBUG: 2015:04:26 23:41:31 3692 MainThread radical : [INFO ] python.interpreter version: 2.7.5+ (default, Sep 17 2013, 15:31:50) [GCC 4.8.1] But we would still be interested to know, of course, what caused your troubles, so please let us know if you have any details on how you installed the new radical.pilot release (and dependencies). Best, Andre. — |
|
Please try the attached script to see if you can reproduce the bug by running it “source FreshRun new”. I found that, in the script:
a. Comment out the last statement “sh launcher_amber_2d.sh”; and run the last statement separately à works b. Include the last statement “sh launcher_amber_2d.sh” à sometime works, sometime doesn’t. I tried on two separate machines and got the same results. Maybe you guys can run and try. (Remember to change the directory setting) Taisung From: Taisung Lee [mailto:taisung@gmail.com] Antons seemed to identity the problem: the problem will occur if I run “python setup.py install” for each module, but will be OK if I use “pip install --upgrade .” instead. You may try if you can reproduce it. Taisung From: Andre Merzky [mailto:notifications@github.com] Taisung, if you continue to have trouble with the current release and if that stop you from doing work, you may want to revert to the previous version: $ virtualenv ve should give you (with RADICAL_VERBOSE, SAGA_VERBOSE and RADICAL_PILOT_VERBOSE all set to DEBUG: 2015:04:26 23:41:31 3692 MainThread radical : [INFO ] python.interpreter version: 2.7.5+ (default, Sep 17 2013, 15:31:50) [GCC 4.8.1] But we would still be interested to know, of course, what caused your troubles, so please let us know if you have any details on how you installed the new radical.pilot release (and dependencies). Best, Andre. — |
|
I'm afraid I don't see the script attached, at least not on the ticket on #23 (comment) . Could you please send it by mail, or paste it if it is not too long? Thanks! |
|
If you do: modify amber_input.json, e.g. allocation, resource etc. everything should work. There is a bug in feature/2d-prof branch (will be fixed soon) but for small number of replicas you should not experience any problems. For your script it should be sufficient to remove: since 'python setup.py install' in RepEx installs radical pilot and it's dependencies anyway, so I don't see any use for that loop. If you want to re-install everything just do; |
|
Original script is: |
|
I can reproduce the problem with: which for the same simulation would give: |
|
Hi Antons, Taisung From: Antons [mailto:notifications@github.com] If you do: rm -rf $HOME/myenv; modify amber_input.json, e.g. allocation, resource etc. sh ./launcher_amber_2d.sh everything should work. There is a bug in feature/2d-prof branch (will be fixed soon) but for small number of replicas you should not experience any problems. For your script it should be sufficient to remove: for module in "radical.utils" "saga-python" "radical.pilot" since 'python setup.py install' in RepEx installs radical pilot and it's dependencies anyway, so I don't see any use for that loop. If you want to re-install everything just do; rm -rf $HOME/myenv; — |
OK, I will.
I don't understand what you mean. All I am trying to say is there is no need to install rp saga-python and utils separately if all these are from master branch. |
|
Taisung, I have slightly modified your script (should not affect any behavior) and it worked fine for me: |
The text was updated successfully, but these errors were encountered: