search

cmbi / kmad-web

1 stars 1 forks source link

time duration #38

Closed cbaakman closed 6 years ago

cbaakman commented 6 years ago

For usage in hope, it is important to know how long a kmad job could possibly take to complete. The presumed maximum is 2 hours. However, it took longer for the following parameters:

sequence: MAAALLARASGPARRALCPRAWRQLHTIYQSVELPETHQMLLQTCRDFAEKELFPIAAQVDKEHLFPAAQVKKMGGLGLLAMDVPEELGGAGLDYLAYAIAMEEISRGCASTGVIMSVNNSLYLGPILKFGSKEQKQAWVTPFTSGDKIGCFALSEPGNGSDAGAASTTARAEGDSWVLNGTKAWITNAWEASAAVVFASTDRALQNKGISAFLVPMPTPGLTLGKKEDKLGIRGSSTANLIFEDCRIPKDSILGEPGMGFKIAMQTLDMGRIGIASQALGIAQTALDCAVNYAENRMAFGAPLTKLQVIQFKLADMALALESARLLTWRAAMLKDNKKPFIKEAAMAKLAASEAATAISHQAIQILGGMGYVTEMPAERHYRDARITEIYEGTSEIQRLVIAGHLLRSYRS

output-type: predict

prediction-methods: d2p2

@joannalange do you have an estimate for the general amount of time a job can take?

joannalange commented 6 years ago

How long did it take? It's because the default prediction source is the d2p2 database, and if the sequence is not there it falls back to running predictors itself (iupred and psipred). The psipred predictor is the reason for such a long runtime. Only thing I could do is increase the numbers of cores that it's running on, or change the predictor to a fast crappy one, however I don't think that's the way to go.

And all the good predictors have such a long runtime so changing the predictor doesn't seem like a viable option.

cbaakman commented 6 years ago

The job was started 2 hours and 20 minutes ago. It hasn't completed yet at the time of writing.

Also, the job id was 'b7b1a8ad-062a-4940-ac04-2aec15021008'.

joannalange commented 6 years ago

I've increased the number of cores and it took around ~10 minutes now Though I think I killed it on chelonium by accident, cause I ran another instance on the same port and then when I killed it seems like the first instance broke... Sorry for that

joannalange commented 6 years ago

And another thing: where can I access the logs from the live version?

cbaakman commented 6 years ago

OK, then I suppose that 2 hours is enough.

joannalange commented 6 years ago

Ah ok, silly - I've made the commit with increasing number of cores on my fork instead of here... But it's fixed now, commit fea2814

cbaakman commented 6 years ago

logs are accessible by running either: docker logs kmadweb_web_1 docker logs kmadweb_celery_1 docker logs kmadweb_databanks_1