Hi
This is just a status update on what I am doing on the topic-libqb branch.
--- CTS ---
I currently have 4 tests commented out (all others pass):
1] CpgMsgOrderBasic
2] CpgMsgOrderZcb
- The issue here is qb_ipcs_event_send() returns try again as the ringbuffer
is full. Steven remember the flow control TODO!
Can we come up with a plan for this?
I can add a memory buffer (like we have in master) in ipc_glue.c
But then we also need an event that goes into totem to adjust the
flowcontrol in the token. Do you have have any other ideas on
how to approach this?
3] WDOnForkBomb
4] RebootOnHighMem
- These work but the tests are not recovering nicely (same reason the
currrent
buildbot is "failing").
a) Both (correctly) cause the VM to reboot
b) I then wait for them to be up (with "ping -nq -c10 -w10 <node>")
ping succeeds
c) then CTS will do "ssh <command>" - this seems to lock up, not sure why.
--- SCHED_RR ---
The other thing is corosync-v2 does not work properly with SCHED_RR on a single
node
so I have it commented it out in the code. I am not sure what to do about this
right now
but be aware. I can look into this a bit further as I get time.
--- totem queue len ---
Remember this hardcoded hack I put in:
static void check_q_level(struct totempg_group_instance *instance)
{
int32_t old_level;
int32_t percent_used = 0;
old_level = instance->q_level;
percent_used = 100 - (totemmrp_avail () * 100 / 800); /*(1024*1024/1500)*/
I remember you (Steven) saying that you would take care of this ^^^ as
the define was incorrect and you didn't like how the queue size
was calculated? Or I can handle this if you can tell me what you were planning.
Regards
-Angus
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais
