[ https://issues.apache.org/jira/browse/YARN-3655?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14557003#comment-14557003 ]
Karthik Kambatla commented on YARN-3655: ---------------------------------------- Comments on the patch: # okToUnreserve ## It was a little hard to wrap my head around. Can we negate it and call it {{isValidReservation(FSSchedulerNode)}}? ## Can we get rid of the if-else and have a simple {{return hasContainerForNode && fitsInMaxShare && !isOverAMShareLimit}}? # Add an {{if (isValidReservation)}} check in {{FSAppAttempt#reserve}} so all the reservation logic stays in one place? # In {{FSAppAttempt#assignContainer(node, request, nodeType, reserved)}}, ## We can get rid of the fitsInMaxShare check immediately preceding the call to {{reserve}}. ## Given {{if (fitsIn(capability, available))}}-block ends in return, we don't need to put the continuation in else. # While adding this check in {{FSAppAttempt#assignContainer(node)}} might work in practice, it somehow feels out of place. Also, assignReservedContainer could also lead to a reservation? # Instead of calling {{okToUnreserve}}/{{!isValidReservation}} in {{FairScheduler#attemptScheduling}}, we should likely add it as the first check in {{FSAppAttempt#assignReservedContainer}}. # Looks like assign-multiple is broken with reserved-containers. The while-loop for assign-multiple should look at both reserved and un-reserved containers assigned. Can we file a follow-up JIRA to fix this? > FairScheduler: potential livelock due to maxAMShare limitation and container > reservation > ----------------------------------------------------------------------------------------- > > Key: YARN-3655 > URL: https://issues.apache.org/jira/browse/YARN-3655 > Project: Hadoop YARN > Issue Type: Bug > Components: fairscheduler > Affects Versions: 2.7.0 > Reporter: zhihai xu > Assignee: zhihai xu > Attachments: YARN-3655.000.patch, YARN-3655.001.patch, > YARN-3655.002.patch > > > FairScheduler: potential livelock due to maxAMShare limitation and container > reservation. > If a node is reserved by an application, all the other applications don't > have any chance to assign a new container on this node, unless the > application which reserves the node assigns a new container on this node or > releases the reserved container on this node. > The problem is if an application tries to call assignReservedContainer and > fail to get a new container due to maxAMShare limitation, it will block all > other applications to use the nodes it reserves. If all other running > applications can't release their AM containers due to being blocked by these > reserved containers. A livelock situation can happen. > The following is the code at FSAppAttempt#assignContainer which can cause > this potential livelock. > {code} > // Check the AM resource usage for the leaf queue > if (!isAmRunning() && !getUnmanagedAM()) { > List<ResourceRequest> ask = appSchedulingInfo.getAllResourceRequests(); > if (ask.isEmpty() || !getQueue().canRunAppAM( > ask.get(0).getCapability())) { > if (LOG.isDebugEnabled()) { > LOG.debug("Skipping allocation because maxAMShare limit would " + > "be exceeded"); > } > return Resources.none(); > } > } > {code} > To fix this issue, we can unreserve the node if we can't allocate the AM > container on the node due to Max AM share limitation and the node is reserved > by the application. -- This message was sent by Atlassian JIRA (v6.3.4#6332)