Forum Discussion
Bret_McGinnis_1
Nimbostratus
NimbostratusAnomolies with [active_members $pool] and [LB::server addr]
I have an iRule that monitors the number of active nodes and sends out Syslog messages when specific thresholds are reached.
On occasion the [act_members $pool] returns a 0 value when there is one or more active members in the pool (according to the monitor applied). The 0 value is return in the LB_SELECTED event.
On occasion, specifically after the above condition occurs, some requests trigger the LB_FAILED event, even though there are still active members in the pool (according to the monitors applied). [act_members [LB::server pool]] returns a value of 0 and [LB::server addr] returns a null value ("").
I'm I correct to believe that I should not get a value of 0 from [act_members $pool] in the LB_SELECTED event if I have monitors applied and they believe there is at least one active member in the pool?
I'm I correct to believe that if the LB_FAILED event is trigger then I either have no active node in the pool or a request failed after pool selection? If yes, would that mean that I should not have active nodes greater than 0 and get a values of 0 and "" from [act_members [LB::server pool]] and [LB::server addr] respectively.
I'm I making any sense at all.
Regards
- You may simply be seeing a discrepancy between the actual load-balancing and the monitors. The monitors generally only run every 15 or 30 seconds, so it may actually take that much time before they notice a node down. This would potentially explain why you could occasionally be seeing LB_FAILED events as yes that event is triggered either because there are no active nodes or because the connection attempt to the server failed. However, the active_members command is driven by the status as reported by the monitors. Is it possible the monitors are flapping your nodes up and down very quickly? I'm not sure how you are determining the nodes are reported up by the monitors. If it merely through the GUI, then it's certainly possibly the node went down and right back up. You may want to check the logs.
