Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Jun 16 22:23
    YanChii labeled #550
  • Jun 16 22:23
    YanChii milestoned #550
  • Jun 16 22:23
    YanChii labeled #550
  • Jun 16 22:23
    YanChii assigned #550
  • Jun 16 22:23
    YanChii opened #550
  • Jun 16 21:41

    YanChii on master

    fix clean_envs for py2 (compare)

  • Jun 16 20:47
    dependabot[bot] edited #538
  • Jun 16 20:47

    dependabot[bot] on pip

    (compare)

  • Jun 16 20:47
    dependabot[bot] closed #543
  • Jun 16 20:47
    dependabot[bot] commented #543
  • Jun 16 20:46
    dependabot[bot] labeled #549
  • Jun 16 20:46
    dependabot[bot] opened #549
  • Jun 16 20:46

    dependabot[bot] on pip

    Bump ansible from 2.8.8 to 2.9.… (compare)

  • Jun 16 20:46
    dependabot[bot] edited #543
  • Jun 16 20:46
    dependabot[bot] edited #538
  • Jun 16 20:46
    dependabot[bot] synchronize #538
  • Jun 16 20:46

    dependabot[bot] on pip

    Bump jinja2 from 2.10 to 2.11.3… (compare)

  • Jun 16 20:46
    dependabot[bot] edited #538
  • Jun 16 20:46
    dependabot[bot] edited #543
  • Jun 16 20:46
    dependabot[bot] edited #538
FilipFrancis
@FilipFrancis
yeah you need to install a new platform version
this one
platform-20210408T172209Z.tgz
then it works
Jan Poctavek
@YanChii
yes, the platform for upcoming DC release is already out. If you have disk installed node, you can do it by
esdc-platform-upgrade v4.5
FilipFrancis
@FilipFrancis
hmm is V4.5 out?
Jan Poctavek
@YanChii
not yet… but the platform is already out
FilipFrancis
@FilipFrancis
ok cool already running the new platform on one node
toybomb
@toybomb:matrix.org
[m]
on smartos there was a script i was using to update platform on the boot stick
can i use that with the esdc platform somehow?
Jan Poctavek
@YanChii
You can. Thanks to @marcheschi we have the USB upgrade script already in DC. See
bash /opt/erigones/bin/eslib/usb-platform-upgrade.sh
toybomb
@toybomb:matrix.org
[m]
awesome thanks
toybomb
@toybomb:matrix.org
[m]
I just noticed that danube checks cpu requirements only based on running not based on allocated? @YanChii
Jan Poctavek
@YanChii
all VMs count as used CPUs. It doesn't matter if the VM is running/stopped/defined. However you can edit node's advanced properties and increase CPU (or memory/disk) ratio to be able to overcommit resources
toybomb
@toybomb:matrix.org
[m]
I see
I don't really think it's logical though
It's not really how other services handle it
Jan Poctavek
@YanChii
depends on what you want to achieve… if you want to run a cloud and present the customers/consumers your compute resources, they will not see your HW nodes, just compute resources you dedicate to them
this way you can optimize your HW utilization internally without exposing this info to customers
toybomb
@toybomb:matrix.org
[m]
Compute is only allocated when scheduled
Most providers work like that, kubernetes works like that
Jan Poctavek
@YanChii
not in the cloud
if you buy a VM, you pay for it even if it's offline
toybomb
@toybomb:matrix.org
[m]
huh? like who?
you mean aws charges you for stopped instances
sure
but it's not the same concept
Jan Poctavek
@YanChii
I understand your use case. You want to have your resource limit computed from the real resource usage and not from the non-running resources.
It's not how DC currently operates. You need to have the cpu coefficient set.
Jan Poctavek
@YanChii
It's possible to implement it. The mgmt code needs to be enriched by one variable that switches the resource computation behavior. Shouldn't be hard. If someone is interested enough, I can help.
toybomb
@toybomb:matrix.org
[m]
I have some urgent things to take care of I'll try to take a stab at it afterwards
Jan Poctavek
@YanChii
Anyway, this would change behavior from cloud-like to vmware-like.
ok, let me know then and I can show you where you need to enrich the code
toybomb
@toybomb:matrix.org
[m]
I actually think you're wrong. I don't think that the allocated compute is actually fully allocated
they kick spot instances off when you start your normal instance
I think that means that theyll happily use the compute but reallocate it to you because you're a paying customer
so yes you pay for it, and yet you'll get your paid for compute when you want to
but it isn't as simple as they won't use that compute at all
maybe then there should be a spot instance flag danube
but that's harder to implement i think
Jan Poctavek
@YanChii
I can explain you the current functionality then and we'll see what can be done.
toybomb
@toybomb:matrix.org
[m]
ok sounds good
not this week though
Paolo Marcheschi
@marcheschi
Hi, I have my mon01 that is getting bigger, is there some command or script to reduce the DB occupation ? Thank you
Jan Poctavek
@YanChii
the DB should do autovacuum… but you can also force it. Search for postgres vacuum.
Paolo Marcheschi
@marcheschi
@YanChii thankyou, I tried but it stopped with no space left on device, so I enlarged the disk and now I'm trying to do that again, the problem is that I need a way to limit the grow of the DB
yanchii
@yanchii:matrix.org
[m]
set lower retention of data… afaik default is 365 days
alternatively you can increase delay between checks so you have fewer samples to store
Paolo Marcheschi
@marcheschi
ok now I have 30 GB, let's see if it is enough, thank you