D4Science Infrastructure

New nodes added to the iptable rules on workspace-repository.d4science.org
(and the new workspace-repository-prod.d4science.org for the future)

The machines work very well and I was able to use them to run 3500 taxonomic searches in a very large repository. I'm just testing the dataminer1-2-genericworker machines that seem to have issues. For sure they are downsized, due also to the scheduled updater of R packages, but it seems they have other issues too with Java libraries. Is it possible to increase their memory configuration and reinstall them?

Gianpaolo Coro wrote:

I'm just testing the dataminer1-2-genericworker machines that seem to have issues. For sure they are downsized, due also to the scheduled updater of R packages, but it seems they have other issues too with Java libraries. Is it possible to increase their memory configuration and reinstall them?

I don't see why a reinstall should be useful. Their resources can be increased, but the distribution and the configurations are up to date with all the other dataminer installations. I don't know if there's some outdated algorithm, but it's not something that a reinstall can solve.

For the time being, if the GARR dataminers are sufficient, the local generic workers can be removed from the haproxy pool and maybe dismissed.

@gianpaolo.coro@isti.cnr.it any opinion on my proposal?

Yes, I have run some tests and the Garr resources are sufficient.

I just removed dataminer[1:2]-genericworkers.d4science.org from the genericworkers backend. Now we have two free dataminer VMs. We can add them to the genericworkers_proto backend or we could destroy them and add more resources to dataminer3-genericworkers.d4science.org. @gianpaolo.coro@isti.cnr.it, your call.

Please, add resources to dataminer3-genericworkers.d4science.org

Gianpaolo Coro wrote:

Please, add resources to dataminer3-genericworkers.d4science.org

Is it a problem if I increase the dataminer1-genericworkers.d4science.org's resources and put it into the genericworkers_proto backend? It lives in an hypervisor with more free memory than the one that hosts dataminer3-genericworkers.d4science.org