How frequently are "low priority workstations" restarted
Hi,
Are there stats or personal experience about how often low-priority workstations are restarted when using dx run cloud_workstation <etc> --priority=low?
Over this weekend, I tried running ~20 workstations with low priority, each of them taking between 4 and 6 hours to do the desired processing, and they were all restarted (some of them 5 times). None of them could finish their tasks before being restarted.
I just wanted to know if this weekend was especially busy or if this is usual.
Thanks
Comments
1 comment
Hello Fidel,
For interactive jobs like Jupyter Lab, Cloud workstation or R it is highly recommended to use high priority which uses only on-demand instance and thus avoids its interruption.
https://dnanexus.gitbook.io/uk-biobank-rap/working-on-the-research-analysis-platform/managing-job-priority
You can use low priority in case you have some brief job, as the time to spot instance interruption is very random.
You can try to use amazon spot instance advisor which can provide you with up to date probability of spot instance interruption for specific instance types.
https://aws.amazon.com/ec2/spot/instance-advisor/
Please sign in to leave a comment.