Page MenuHomePhabricator

Audit hyperthreading on analytics nodes.
Closed, ResolvedPublic


Hyperthreading does not seem to be consistently set across hadoop worker and kafka nodes. All of these nodes have 12 physical cores, and, here is what the OS thinks they have:

analytics1011.eqiad.wmnet: 24
analytics1012.eqiad.wmnet: 12 (kafka)
analytics1013.eqiad.wmnet: 24
analytics1014.eqiad.wmnet: 24
analytics1015.eqiad.wmnet: 24
analytics1016.eqiad.wmnet: 24
analytics1017.eqiad.wmnet: 24
analytics1018.eqiad.wmnet: 24 (kafka)
analytics1019.eqiad.wmnet: 24
analytics1020.eqiad.wmnet: 24
analytics1021.eqiad.wmnet: 12 (kafka)
analytics1022.eqiad.wmnet: 12 (kafka)
analytics1028.eqiad.wmnet: 12
analytics1029.eqiad.wmnet: 12
analytics1030.eqiad.wmnet: 12
analytics1031.eqiad.wmnet: 12
analytics1032.eqiad.wmnet: 12
analytics1033.eqiad.wmnet: 12
analytics1034.eqiad.wmnet: 12
analytics1035.eqiad.wmnet: 12
analytics1036.eqiad.wmnet: 12
analytics1037.eqiad.wmnet: 12
analytics1038.eqiad.wmnet: 12
analytics1039.eqiad.wmnet: 12
analytics1040.eqiad.wmnet: 12
analytics1041.eqiad.wmnet: 12

Event Timeline

Ottomata claimed this task.
Ottomata raised the priority of this task from to Low.
Ottomata updated the task description. (Show Details)
Ottomata added a project: Analytics-Clusters.
Ottomata added a subscriber: Ottomata.

I will be starting on nodes 1028 - 1041 today.

Hyperthreading is now turned on all Hadoop nodes.

@Ottomata can we close this task or do you want to follow up more? (Not sure if having an automated script that polls ht usage and alerts if not set would be worth the time)