Open topic with navigation
This sections lists ideas for monitoring and tweaking your database resource usage.
mem(default 2GB). This is the memory threshold and is used as an advisory target for the total memory in use by a single NuoDB process (TE or SM). See Database Options for more information.
nuodb_systemdatabase processes, or adjust the property
metricsPurgeAge(see Host Properties (default.properties)), to purge old statistics more often. Otherwise, you may incur performance issues because the
nuodb_systemdatabase processes are handling too much data.
NuoDB recommends that you set a host alarm on memory usage to warn if the database reaches 80% usage. You can do this using the
nuodbmgr utility as shown below. This example creates an
alarmDefinition and names the alarm
mem-usage-alarm. It specifies usage of the
OS-memUsedPercent metric and indicates that if this metric remains above 80% for ten minutes for the entire domain then the alarm should be triggered. The severity level is
Warning and the action is a notification event.
nuodbmgr --broker hostname --password password --command "create alarmDefinition metric name 'mem-usage-alarm' description 'warn if the host reaches 80% usage' metricName OS-memUsedPercent dimension Domain aggrFunc Average operator GE threshold 80 breachDuration 10 action NotifyEvent severity Warning"Created Alarm Definition [mem-usage-alarm]: type=StatAlert dimension=Domain entity=* metric=OS-memUsedPercent(Average) GE 80.00 600s (Warning, Reset)
Refer to the Setting Up Failure Detection for information regarding failure detection. Note that on AWS EC2 instances, failure detection scenarios are hampered because sockets take a while to be cleared, so running without failure detection can be particularly problematic.
If you plan on saving the EC2 instance as an Amazon Machine Image (AMI) from which to clone other instances, then do not start the broker or agent process. If you do, it will create a "stableId" in the
/var/opt/nuodb folder and in the durable domain configuration. If you then later try to peer the two machines together, they will conflict because of the duplicate stableId. If you do start the broker/agent for testing purposes, just wipe the system clean before snapshotting the machine and especially delete files under