IGMP group member node unexpectedly dropped from the igmp group list
Issue
Customer is using igmp multicast packets to spread trading bets/offers throughout his trading environment.
From time to time it happened that some trading node was kicked out of igmp group receiving trading data till refresh timer expired
and the igmp group members list was refreshed.
In such cases actual offers and bits could not be seen in such trading node till the refresh of the igmp group member list.
This caused unacceptable trading delays in customers trading environment.
It has been identified that customer igmp trading environment contains more than usual one igmp querier.
These igmp queriers together produced igmp group member requests in higher rate than maximal igmp group member response timer is defined.
This means that new igmp group member query could arrive sooner than response to the previous one was sent by the igmp group member trading node.
Unfortunately there was a bug in rhel igmp group member code handling igmp group member queries.
This bug caused that when new igmp group member query arrived to some igmp group member node the timer
for actual response was scheduled with no respect of possibly pending previous igmp group member query.
This bug together with more than one igmp queriers producing queries in fast rate caused that the response
from igmp group member node was never sent and the node was released from igmp group member list maintained by switch.
This caused that this node was out of distribution list for the igmp group and was missing data till the refresh of the list done by the switch.
Environment
All RHEL versions till RHEL-7.4
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
