benchmark scenarios with high "times" can result in 'MySQL server has gone away'
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Rally |
Invalid
|
Undecided
|
Unassigned |
Bug Description
Using the follow scenario as a task:
{
"Ceilometer
{
"args": {
},
},
}
}
}
]
}
The traceback below happens. The error happens after all ceilometer api requests have been made, so presumably during the results processing phase.
Exception in thread Thread-10:
Traceback (most recent call last):
File "/usr/lib64/
self.run()
File "/usr/lib64/
self.
File "/opt/stack/
"scenario_
File "/opt/stack/
db.
File "/opt/stack/
return IMPL.task_
File "/opt/stack/
result.save()
File "/opt/stack/
super(
File "/usr/lib/
session.flush()
File "/usr/lib64/
self.
File "/usr/lib64/
transaction
File "/usr/lib64/
compat.
File "/usr/lib64/
transaction
File "/usr/lib64/
transaction
File "/usr/lib64/
t[1].rollback()
File "/usr/lib64/
self.
File "/usr/lib64/
self.
File "/usr/lib64/
self.
File "/usr/lib64/
util.
File "/usr/lib64/
reraise(
File "/usr/lib64/
self.
File "/usr/lib64/
dbapi_
DBConnectionError: (OperationalError) (2006, 'MySQL server has gone away') None None
Changed in rally: | |
status: | Incomplete → Confirmed |
From:
5:59pm] msdubov_: cdent Hi, thanks for the report. Has this bug appeared in your machine right after Rally installation or after some time you used it?
[6:00pm] cdent: I update my rally code to master earlier today and have run several tests throughout the day
[6:00pm] cdent: when the "times" is less than 2000 or so, it doesn't happen
[6:00pm] cdent: up around 5000 it does
[6:01pm] cdent: I initially assumed it was a concurrency problem, but it appears to have more to do with volume of data.
[6:01pm] msdubov_: cdent Quite interesting, could you please add that as a comment to that bug?
Is there a huge query happening within mysql that is perhaps allowing the connection to time out?