Page MenuHomePhabricator

Tear down the datahubsearch service and its constituent VMs
Closed, ResolvedPublic

Description

As part of T402408: Upgrade Datahub to a more recent version we migrated the OpenSearch cluster behind datahub.wikimedia.org to Kubernetes, running on the dse-k8s-eqiad cluster.

Because of this, we no longer need the datahubsearch service that is provided by the three Gantei virtual machines: datahubsearch100[1-3].eqiad.wmnet

This ticket will track the decommissioning of the service and the VMs.

We need to remove datahubsearch from at least one OpenSearch dashboard as well.

Event Timeline

Icinga downtime and Alertmanager silence (ID=a24bd063-e76d-49af-86b5-aec247d1d3d6) set by bking@cumin2003 for 4:00:00 on 3 host(s) and their services with reason: Using datahubsearch1001 to test bookworm reimages

datahubsearch[1001-1003].eqiad.wmnet

Icinga downtime and Alertmanager silence (ID=21dd99da-77d1-4c0e-9bf2-e6fc5c0efd56) set by bking@cumin2003 for 7 days, 0:00:00 on 2 host(s) and their services with reason: Using datahubsearch1001 to test bookworm reimages

datahubsearch[1002-1003].eqiad.wmnet
bking changed the task status from Open to In Progress.Wed, Jul 15, 10:36 PM
bking claimed this task.
bking triaged this task as Low priority.

Change #1311530 had a related patch set uploaded (by Bking; author: Bking):

[operations/puppet@production] datahubsearch: move to 'insetup' role

https://gerrit.wikimedia.org/r/1311530

Change #1311533 had a related patch set uploaded (by Bking; author: Bking):

[operations/puppet@production] datahubsearch: remove references to Puppet plans

https://gerrit.wikimedia.org/r/1311533

Change #1311538 had a related patch set uploaded (by Bking; author: Bking):

[operations/dns@master] datahubsearch: remove DNS record

https://gerrit.wikimedia.org/r/1311538

Change #1311542 had a related patch set uploaded (by Bking; author: Bking):

[operations/alerts@master] opensearch: remove references to soon-to-be-decom'd service

https://gerrit.wikimedia.org/r/1311542

Change #1311542 merged by Bking:

[operations/alerts@master] opensearch: remove references to soon-to-be-decom'd service

https://gerrit.wikimedia.org/r/1311542

Change #1311538 merged by Bking:

[operations/dns@master] datahubsearch: remove DNS record

https://gerrit.wikimedia.org/r/1311538

Change #1311899 had a related patch set uploaded (by Bking; author: Bking):

[operations/puppet@production] datahubsearch: move service from production to lvs_setup

https://gerrit.wikimedia.org/r/1311899

Change #1312517 had a related patch set uploaded (by Bking; author: Bking):

[operations/puppet@production] datahub search: move service to 'service_setup' state.

https://gerrit.wikimedia.org/r/1312517

Change #1311899 merged by Bking:

[operations/puppet@production] datahubsearch: move service from production to lvs_setup

https://gerrit.wikimedia.org/r/1311899

Change #1312517 merged by Bking:

[operations/puppet@production] datahub search: move service to 'service_setup' state.

https://gerrit.wikimedia.org/r/1312517

Change #1311533 merged by Bking:

[operations/puppet@production] datahubsearch: remove references to Puppet plans

https://gerrit.wikimedia.org/r/1311533

Change #1311530 merged by Bking:

[operations/puppet@production] datahubsearch: move to 'insetup' role

https://gerrit.wikimedia.org/r/1311530

cookbooks.sre.hosts.decommission executed by bking@cumin2003 for hosts: datahubsearch[1002-1003].eqiad.wmnet

  • datahubsearch1002.eqiad.wmnet (PASS)
    • Downtimed host on Icinga/Alertmanager
    • Found Ganeti VM
    • VM shutdown
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox
    • Removed from DebMonitor
    • Removed from Puppet server and PuppetDB
    • VM removed
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox
  • datahubsearch1003.eqiad.wmnet (PASS)
    • Downtimed host on Icinga/Alertmanager
    • Found Ganeti VM
    • VM shutdown
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox
    • Removed from DebMonitor
    • Removed from Puppet server and PuppetDB
    • VM removed
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox

datahubsearch1001 (which had fallen out of the PuppetDB ) is still in process of decommissioning:

[bking@ganeti1048] ~$ sudo gnt-instance list | grep datahubsearch
datahubsearch1001.eqiad.wmnet       kvm        debootstrap+default ganeti1057.eqiad.wmnet ADMIN_down      -

Assuming the decom cookbook finishes cleanly, this should be the last step in getting rid of the cluster.

cookbooks.sre.hosts.decommission executed by bking@cumin2003 for hosts: datahubsearch1001.eqiad.wmnet

  • datahubsearch1001.eqiad.wmnet (WARN)
    • Host not found on Icinga, unable to downtime it
    • Found Ganeti VM
    • VM shutdown
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox
    • Removed from DebMonitor
    • Removed from Puppet server and PuppetDB
    • VM removed
    • Started forced sync of VMs in Ganeti cluster eqiad to Netbox
bking closed this task as Resolved.EditedWed, Jul 22, 2:58 PM

Per the above message, all VMs and their associated configs have been removed. As such, I'm closing this ticket. However, do feel free to reopen if we missed anything.

I still need to remove the hostnames from Puppet's site.pp.

Change #1313978 had a related patch set uploaded (by Bking; author: Bking):

[operations/puppet@production] datahubsearch: remove references to recently-decommed cluster

https://gerrit.wikimedia.org/r/1313978

Change #1313978 merged by Bking:

[operations/puppet@production] datahubsearch: remove references to recently-decommed cluster

https://gerrit.wikimedia.org/r/1313978

OK, I believe that was the final step. Closing...