Investigate if Node Managers can be restarted without impacting running containers
Closed, ResolvedPublic5 Estimated Story Points
Actions

Assigned To

Authored By

	elukey
	Feb 1 2017, 5:13 PM

Description

Essentially we want to be able to restart the Node Manager daemons on the Hadoop worker nodes without impacting the running containers.

	Subject	Repo	Branch	Lines +/-
	Update the cdh's module sha	operations/puppet	production	+1 -1
	Enable Yarn's Node Manager recovery to allow graceful restarts	operations/puppet/cdh	master	+5 -0

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptFeb 1 2017, 5:13 PM

elukey triaged this task as High priority.Feb 1 2017, 5:14 PM

Change 336203 had a related patch set uploaded (by Elukey):
Enable Yarn's Node Manager recovery to allow graceful restarts

Mentioned in SAL (#wikimedia-operations) [2017-02-06T13:30:05Z] <elukey> applied https://gerrit.wikimedia.org/r/#/c/336203/ manually to analytics1028 (hadoop worker node) as live test - T156932

Milimetric set the point value for this task to 5.Feb 6 2017, 4:42 PM

Change 336203 merged by Elukey:
Enable Yarn's Node Manager recovery to allow graceful restarts

Change 336380 had a related patch set uploaded (by Elukey):
Update the cdh's module sha

Change 336380 merged by Elukey:
Update the cdh's module sha

Mentioned in SAL (#wikimedia-operations) [2017-02-07T14:19:50Z] <elukey> restarting all the Yarn Node Managers on the Hadoop worker nodes to pick up the new config - T156932

• Nuria closed this task as Resolved.Feb 14 2017, 4:05 PM