FM messaging socket reads that are triggered by FM API calls from
client services have been seen to rarely but occasionally block/stall
the fmManager process. This fmManager stall can then lead to other
client service process stalls which in the case of mtcAgent has been
seen to lead to uncontrolled switch of activity ; aka Swact.
This update adds a 5 second socket read timeout to FM's client services
socket setup to avoid the prolonged blocking cases that lead to Swact
or adversely affect (block) other client service process execution.
Setting a read timeout on Linux sockets is a good programming practice.
Doing so it helps ensure that an application, FM and client services
do not hang indefinitely if a network operation like a socket read
becomes unresponsive.
Configuring a timeout helps manage network communication reliability
and efficiency, especially in applications where responsiveness is
critical. Especially in server-client application such as FM.
Test Plan:
PASS: Verify AIO DX system install.
PASS: Verify blocked socket timeout and error log after 5 seconds.
PASS: Verify unblocked socket reads complete successfully.
PASS: Verify alarm assert/clear functions operate normally.
PASS: Verify set socket timeout failure handling.
PASS: Verify fmManager is not leaking files or memory.
PASS: Verify rook-ceph apply remove 100 loop soak
- no stall or swact
- AIO DX
- with 2 OSDs on each controller
Closes-Bug: 2088025
Change-Id: I1d947bccf9faeedcc2b96c7bc398fbab77b7ae09
Signed-off-by: Eric MacDonald <eric.macdonald@windriver.com>
54f9fed7c3