*** factor has joined #openvswitch | 00:20 | |
zhouhan | avishnoi: the reason is the timer impacts leader election timeout. If it is changed dramatically, the current heart-beat can get timed-out while the timer is being propagated to the followers, and unnecessary disturbance in the cluster will be triggered. | 00:47 |
---|---|---|
avishnoi | zhouhan: but won’t master commit the change last, before propogating it to the followers? | 00:51 |
zhouhan | avishnoi: yes, leader always applies the change first. So if you are enlarging the timer very long, follower may still expecting a short timeout, but leader may started waiting for a long time to send next heartbeat | 00:58 |
zhouhan | It wouldn't take too many times of change with power of 2 to set it to the desired value, but of course it would be better to have the capability to set it directly at cluster initialization. It is on my list but just not a priority :) | 01:03 |
avishnoi | zhouhan: okay, because leader append the log entry first, it start acting on it. | 01:04 |
zhouhan | yep | 01:05 |
avishnoi | zhouhan: in case of no client initiated transaction, periodic heart beat is sent before every election-timer ? or periodic heartbeat works on diffrent counter ? | 01:05 |
zhouhan | every 1/3 of election timer | 01:06 |
zhouhan | roughly. There is a random delta added | 01:07 |
avishnoi | zhouhan: i think if heart beat can use different counter, we can update election timer to any value without issue. | 01:10 |
avishnoi | zhouhan: because providing this timer at the time of initialization won’t help if user change the timer dynamically | 01:10 |
avishnoi | zhouhan: user can still hit the same issue. | 01:10 |
zhouhan | ok, but would the x2 limit be a problem still? | 01:11 |
avishnoi | zhouhan: not really, it’s just a minor operational inconvinience, nothing else. | 01:12 |
avishnoi | zhouhan: admin will have to cook-up this logic explictly, rather than simply firing one command. | 01:13 |
avishnoi | zhouhan: But nothing major as such i think. | 01:13 |
*** dcbw has quit IRC | 01:14 | |
zhouhan | I think if initial value can be set properly, there shouldn't be any operational problem, right? Unless the initial value is set to an insane value. | 01:17 |
zhouhan | Our practice now is simply adjust the value to desired right after a new cluster is initialized. This doesn't happen very often in reality. | 01:18 |
zhouhan | Then during special operations we usually only need to run one command to adjust it to the new desired value. | 01:19 |
zhouhan | avishnoi: It would be great if there is a way to improve. I am not quite clear about the idea of "different counter". Maybe we can discuss in ML with more details. Patches are welcome as well :) | 01:21 |
*** dholler has quit IRC | 02:33 | |
*** dholler has joined #openvswitch | 02:45 | |
*** factor has quit IRC | 02:49 | |
*** acidfu has quit IRC | 03:14 | |
*** factor has joined #openvswitch | 03:15 | |
*** zhouhan has quit IRC | 03:16 | |
*** zhouhan has joined #openvswitch | 03:17 | |
*** zhouhan_ has joined #openvswitch | 03:18 | |
*** zhouhan has quit IRC | 03:21 | |
avishnoi | zhouhan_: Well setting this election timer depends on the environment (Latency + Disk speed + CPU load on process running raft implementation etc). So you might have to keep tuning in how things are changing in your deployment environment. It can be improved, but as i said, it’s not really critical. | 03:39 |
*** psahoo has joined #openvswitch | 05:03 | |
*** anilvenkata has joined #openvswitch | 05:17 | |
*** yamamoto has quit IRC | 06:08 | |
*** mmirecki has joined #openvswitch | 06:24 | |
*** yamamoto has joined #openvswitch | 06:36 | |
*** mbarroso has joined #openvswitch | 07:01 | |
*** eelco has joined #openvswitch | 07:10 | |
*** slaweq_ has joined #openvswitch | 07:11 | |
*** yamamoto has quit IRC | 07:20 | |
*** yamamoto has joined #openvswitch | 07:25 | |
*** dceara has joined #openvswitch | 07:27 | |
*** dceara has quit IRC | 07:33 | |
*** dceara has joined #openvswitch | 07:44 | |
*** links has joined #openvswitch | 08:16 | |
*** thaller has joined #openvswitch | 08:21 | |
*** thaller_ has quit IRC | 08:23 | |
*** timothy has joined #openvswitch | 08:29 | |
*** zhouhan_ has quit IRC | 09:30 | |
*** zhouhan has joined #openvswitch | 09:30 | |
*** thaller_ has joined #openvswitch | 09:45 | |
*** thaller has quit IRC | 09:48 | |
*** lynxis has quit IRC | 09:59 | |
*** lynxis has joined #openvswitch | 10:01 | |
*** matteo| is now known as matteo | 10:03 | |
*** slaweq_ is now known as slaweq | 10:43 | |
*** links has quit IRC | 12:17 | |
*** bostondriver has joined #openvswitch | 12:18 | |
*** links has joined #openvswitch | 12:18 | |
*** rcernin has quit IRC | 13:14 | |
*** _lore_ has quit IRC | 13:17 | |
*** thaller_ has quit IRC | 13:27 | |
*** thaller has joined #openvswitch | 13:32 | |
*** _lore_ has joined #openvswitch | 13:36 | |
*** dholler has quit IRC | 13:57 | |
*** acidfu has joined #openvswitch | 14:05 | |
*** dholler has joined #openvswitch | 14:10 | |
*** yamamoto has quit IRC | 14:23 | |
*** yamamoto has joined #openvswitch | 14:41 | |
*** yamamoto has quit IRC | 14:42 | |
*** yamamoto has joined #openvswitch | 14:42 | |
*** yamamoto has quit IRC | 14:42 | |
*** yamamoto has joined #openvswitch | 14:43 | |
*** yamamoto has quit IRC | 14:44 | |
*** yamamoto has joined #openvswitch | 14:44 | |
*** yamamoto has joined #openvswitch | 14:44 | |
*** acidfu has quit IRC | 14:47 | |
*** yamamoto has quit IRC | 14:49 | |
*** kevinbenton has quit IRC | 15:05 | |
*** kevinbenton has joined #openvswitch | 15:06 | |
*** atpa8a has quit IRC | 15:07 | |
*** mmirecki has quit IRC | 15:18 | |
*** dholler has quit IRC | 15:36 | |
*** balkamos has quit IRC | 15:40 | |
*** balkamos has joined #openvswitch | 15:41 | |
*** atpa8a has joined #openvswitch | 15:46 | |
*** tbachman has quit IRC | 15:53 | |
*** yamamoto has joined #openvswitch | 15:54 | |
*** tbachman has joined #openvswitch | 15:56 | |
*** ihrachys has joined #openvswitch | 16:06 | |
*** ihrachys has quit IRC | 16:09 | |
*** ihrachys has joined #openvswitch | 16:09 | |
*** eelco has quit IRC | 16:17 | |
*** dholler has joined #openvswitch | 16:17 | |
*** acidfu has joined #openvswitch | 16:23 | |
*** kklimonda has quit IRC | 16:28 | |
*** kklimonda has joined #openvswitch | 16:30 | |
*** dholler has quit IRC | 16:31 | |
*** jraju__ has joined #openvswitch | 16:32 | |
*** links has quit IRC | 16:33 | |
*** dcbw has joined #openvswitch | 16:34 | |
*** d1rewolf has quit IRC | 16:37 | |
*** edsousa has quit IRC | 16:38 | |
*** kklimonda has quit IRC | 16:38 | |
*** d1rewolf has joined #openvswitch | 16:39 | |
*** edsousa has joined #openvswitch | 16:39 | |
*** kklimonda has joined #openvswitch | 16:39 | |
*** fkautz has quit IRC | 16:40 | |
*** jamespage has quit IRC | 16:40 | |
*** d1rewolf has quit IRC | 16:40 | |
*** fkautz has joined #openvswitch | 16:41 | |
*** jamespage has joined #openvswitch | 16:42 | |
*** jbenet has quit IRC | 16:43 | |
*** d1rewolf has joined #openvswitch | 16:44 | |
*** edsousa has quit IRC | 16:44 | |
*** jbenet has joined #openvswitch | 16:45 | |
*** d1rewolf has quit IRC | 16:46 | |
*** d1rewolf has joined #openvswitch | 16:47 | |
*** edsousa has joined #openvswitch | 16:47 | |
*** d1rewolf has quit IRC | 16:49 | |
*** d1rewolf has joined #openvswitch | 16:49 | |
*** mmirecki has joined #openvswitch | 16:52 | |
*** fbl has quit IRC | 16:57 | |
*** jraju__ has quit IRC | 16:57 | |
*** aserdean has joined #openvswitch | 17:01 | |
aserdean | Hi all | 17:01 |
*** mmirecki has quit IRC | 17:04 | |
*** inflatador has joined #openvswitch | 17:12 | |
*** mbarroso has quit IRC | 17:13 | |
*** mmirecki has joined #openvswitch | 17:17 | |
*** mmirecki has quit IRC | 17:30 | |
*** teardown has quit IRC | 17:33 | |
*** teardown has joined #openvswitch | 17:33 | |
*** dcbw has quit IRC | 17:33 | |
*** dcbw has joined #openvswitch | 17:42 | |
*** donhw_ has joined #openvswitch | 17:57 | |
*** gmg1 has joined #openvswitch | 17:59 | |
*** donhw has quit IRC | 18:00 | |
*** zhouhan_ has joined #openvswitch | 18:09 | |
*** zhouhan has quit IRC | 18:13 | |
*** psahoo has quit IRC | 18:45 | |
*** rebrec79 has quit IRC | 18:52 | |
*** bern has quit IRC | 18:56 | |
*** atpa8a has quit IRC | 19:02 | |
*** slaweq has quit IRC | 19:14 | |
*** slaweq has joined #openvswitch | 19:20 | |
*** atpa8a has joined #openvswitch | 19:20 | |
*** timothy has quit IRC | 19:25 | |
*** rebrec79 has joined #openvswitch | 19:33 | |
*** mmirecki has joined #openvswitch | 20:07 | |
*** troulouliou_div2 has quit IRC | 20:10 | |
*** mmirecki has quit IRC | 21:05 | |
*** mmirecki has joined #openvswitch | 21:30 | |
*** slaweq has quit IRC | 21:42 | |
*** mmirecki has quit IRC | 21:45 | |
*** inflatador_ has joined #openvswitch | 22:02 | |
*** inflatador has quit IRC | 22:04 | |
*** inflatador_ is now known as inflatador | 22:04 | |
*** dceara has quit IRC | 22:13 | |
*** rcernin has joined #openvswitch | 22:17 | |
acidfu | heya folks | 22:19 |
acidfu | how can I clear this field | 22:19 |
acidfu | sudo ovs-vsctl clear Open_vSwitch . other_config:n-handler-threads | 22:19 |
acidfu | ovs-vsctl: Open_vSwitch does not contain a column whose name matches "other_config:n-handler-threads" | 22:19 |
acidfu | I mean I can do: ovs-vsctl --no-wait set Open_vSwitch . other_config:n-revalidator-threads=25 | 22:20 |
acidfu | set works, but not clear ? :/ | 22:20 |
acidfu | how to I clear it back to empty ? | 22:20 |
acidfu | thanks ! | 22:20 |
acidfu | I want to get back to this: | 22:27 |
acidfu | # ovs-vsctl get Open_vSwitch . other-config | 22:27 |
acidfu | {} | 22:27 |
*** rebrec79 has quit IRC | 22:32 | |
*** rebrec79 has joined #openvswitch | 22:33 | |
*** inflatador has quit IRC | 23:49 | |
*** inflatador has joined #openvswitch | 23:56 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!