It seems that the poll call waits only for 10000ms=10sec , but the call takes over 20-30sec to complete.
13:22:38.269029 poll([{fd=3, events=POLLIN}], 1, 10000) = 0 (Timeout) <10.010702>
Manual curl -v http://169.254.169.254/latest/meta-data takes over 20sec .
mysql> show processlist;
show around 2k connections
default max limit is 4096
We increased the api & conductor worker from 8 to 16, but still no change.
If we test the curl within 5sec of all service restart, the curl replies in 2sec.
This environment has around 75 overcloud nodes.
We already have a bug for providing better WSGI process/thread tuning for keystone on the undercloud. Marking this as a duplicate of our OSP 10.z clone for this issue.
*** This bug has been marked as a duplicate of bug 1523438 ***
It seems that the poll call waits only for 10000ms=10sec , but the call takes over 20-30sec to complete. 13:22:38.269029 poll([{fd=3, events=POLLIN}], 1, 10000) = 0 (Timeout) <10.010702> Manual curl -v http://169.254.169.254/latest/meta-data takes over 20sec . mysql> show processlist; show around 2k connections default max limit is 4096 We increased the api & conductor worker from 8 to 16, but still no change. If we test the curl within 5sec of all service restart, the curl replies in 2sec. This environment has around 75 overcloud nodes.