High Availability Installation / Suspending ACM |
When an upgrade is performed, ACM should be disabled.
$ /opt/aspera/acm/bin/acmctl –D ACM is disabled globally.
Once ACM is disabled, it will stop performing status file updates and it will transition to the expired state. The services are allowed to continue running.
The Status file: expired output below confirms that ACM has been disabled.
$ /opt/aspera/acm/bin/acmctl -i Checking current ACM status... Aspera Cluster Manager for Orchestrator - status ------------------------------------------ Local hostname: orchestrator1 Active node: orchestrator1 (me) Status of this node: active Status file: expired Disabled globally: yes Disabled on this node: no Database configuration file --------------------------- Database host: localhost Orchestrator active/active services status ------------------------------------ Apache: running Orchestrator Status: -> Orchestrator Engine running with pid: 29500 -> Mongrel serving orchestrator on port 3000 is running with pid: 29590 -> Mongrel serving orchestrator on port 3001 is running with pid: 29593 -> Mongrel serving orchestrator on port 3002 is running with pid: 29596 -> Orchestrator Monitor running with pid: 29615 -> Asynchronous Worker Process 0 is running with pid: 29617 -> Asynchronous Worker Process 1 is running with pid: 29619 -> Synchronous Worker Process 2 is running with pid: 29621 -> Synchronous Worker Process 3 is running with pid: 29624 -> Synchronous Worker Process 4 is running with pid: 29626 -> Synchronous Worker Process 5 is running with pid: 29628 Orchestrator active/passive services status ------------------------------------------ MySQL: stopped
# tail –f /opt/aspera/acm/log/acm4orchestrator.log 2016-10-28 16:48:01 (-0800) acm4orchestrator haorchestrator1 (21470): ACM is now disabled globally: aborting 2016-10-28 16:48:01 (-0800) acm4orchestrator haorchestrator1 (14664): ACM is now disabled globally: aborting
$ /opt/aspera/acm/bin/acmctl –E ACM is enabled globallyOne of the Orchestrator servers will become the active node, with all associated services started, and the other will be passive.
If the active node calls acm4orchestrator before the passive node, it only updates the status file.
If the passive node calls acm4orchestrator before the active node, it does not perform an automatic failover, because the status file transitioned into the current state when ACM was re-enabled.