This seems to be a common issue, but none of the solutions have worked. I tried the REST recipe mentioned in the Apache docs and also restarted the ambari-server. No luck. The presence of these notifications has masked some actual service related issues and I'd really like to clear them.
The advantage is that I can see what truly does need restart. Ever since installing Ranger, the normal restart function has stopped working. I need to go host by host in Ambari and use the restart pulldown for each one individually to actually perform the operation. With everything showing a restart-needed status, it makes it almost impossible in a 32 machine cluster to find real issues. After a day of fighting with it, I've given up on getting the restart to work correctly and I'll settle for simply knowing where to go.
Hello, can you confirm what you are seeing? Is it that on certain hosts, some of the client components are saying a "restart" is needed? Does those hosts also have master and/or slave components, or just clients? What version of Ambari are you using?
It varies. We have an edge (client only) node where just about all clients are marked for restart. In the cluster proper we see this on a large number of machines (but not all) that have both master and slave components. Problem started after installing Ranger. Ambari's automated restart stopped working and I have to go through manually. Would help to clear out the "noise" of the client components (what needs "starting" in a client anyway?).
Ok. I wonder if this is a bug that is surfacing...
I've made a comment on the JIRA to see if this might also affect Ambari 2.1 or 2.2 (and hence, maybe you are hitting this).