-
Title
Fglam: FglAM configuration to store more messages in case of FMS communication loss -
Description
FglAM configuration to store more messages in the event of an FMS outage or maintenance
-
Resolution
The Foglight Agent Manager accumulates messages that are destined to be sent either upstream or downstream in queues between connections. This prevents messages from getting lost in the event of a disconnection.
The 'MaxQueueSize' setting specifies the maximum number of messages that will be stored in the queue. If the queue overflows then the newest messages will be discarded. Zero means messages will never be queued, and -1 means that the queues will be allowed to grow indefinitely.
'MaxDiskSpace' indicates how much disk space (in kB) can be used to store messages. Zero means messages will never be written to disk, and -1 means no limit.
'MaxBatchSize' is the number of messages that will be sent from the queue when a connection has been successfully made. -1 means every available message will be included in each batch.
Changing these settings is only recommended when instructed to do so by Quest Software support. Most FglAM clients will use the default settings.
The upstream verified queue includes additional restrictions on the values. In particular, the 'MaxQueueSize' can never be 0, and the 'MaxDiskSpace' can never be 0.The default settings in the fglam.config.xml file are
<upstream max-queue-size="-1" allow-runtime-change="true"
max-disk-space="102400"
max-batch-size="500"/>
<upstream-verified max-queue-size="-1" allow-runtime-change="true"
max-disk-space="51200"
max-batch-size="250"/>
<downstream max-queue-size="-1" allow-runtime-change="true"
max-disk-space="102400"
max-batch-size="500"/> -
Additional Information
Note: The limitations of extending the duration from the default 1 hour window are that your storage requirements in the fglam disk cache will increase, and the FMS will be subjected to an alarm storm from all the stored data when the FMS resumes. This can place a significant load on the FMS during the restart, but that depends on the number and types of agents whether it's a problem or not for a particular use case.