===========================================
#opendaylight-meeting: Kernel Projects Call
===========================================


Meeting started by rovarga_ at 16:01:50 UTC.  The full logs are
available at
http://meetings.opendaylight.org/opendaylight-meeting/2019/kernel_projects_call/opendaylight-meeting-kernel_projects_call.2019-06-11-16.01.log.html
.



Meeting summary
---------------

* LINK: https://jira.opendaylight.org/browse/NETCONF-626  (jamoluhrsen,
  16:01:56)
* NETCONF-626  (rovarga_, 16:07:57)
* CONTROLLER-1900  (rovarga_, 16:08:02)
* NETCONF-626  (rovarga_, 16:08:51)
  * LINK: https://jira.opendaylight.org/browse/NETCONF-626  (rovarga_,
    16:08:58)
  * we fail to share SchemaContext when devices are being connected
    concurrently  (rovarga_, 16:11:29)
  * the caching is being done in SharedSchemaContextFactory in yangtools
    (rovarga_, 16:14:54)
  * the problem is that while we check the cache before trying to
    assemble a SchemaContext, if we have concurrent access we do not
    recheck if a result was computed  (rovarga_, 16:19:33)
  * this results in classic TOCTOU race  (rovarga_, 16:19:49)
  * which is not harmfull (the system works) but is highly inefficient,
    as the concurrently-computed results are used independently aree not
    deduplicated  (rovarga_, 16:20:21)
  * once a result is loaded though, all subsequent accesses will use the
    last computed result  (rovarga_, 16:20:40)

* CONTROLLER-1900  (rovarga_, 16:21:28)
  * Fluorine SR3 blocker  (rovarga_, 16:21:34)
  * it seems we have the root cause  (rovarga_, 16:21:42)
  * it's not that bad, as it does not affect general CDS users
    (rovarga_, 16:22:01)
  * but it is bad in that it affects EOS  (rovarga_, 16:22:48)
  * LINK:
    https://jenkins.opendaylight.org/releng/job/controller-csit-3node-clustering-ask-all-fluorine/237
    (LuisGomez, 16:24:13)
  * at the end of the day it is an issue where latest state may not be
    visible on followers for up to heartbeat interval  (rovarga_,
    16:26:43)
  * LINK: https://jira.opendaylight.org/browse/CONTROLLER-1900
    (rovarga_, 16:27:09)
  * this is hurtful when the selected EOS owner is on a follower, as its
    mechanics are tied to DTCL on that follower  (rovarga_, 16:27:38)
  * hence there can be a delay until the selected candidate becomes
    operational  (rovarga_, 16:27:55)
  * we have a patch, being validated at the link Luis linked  (rovarga_,
    16:28:14)
  * which should alleviate the problem  (rovarga_, 16:28:23)



Meeting ended at 16:48:50 UTC.



People present (lines said)
---------------------------

* rovarga_ (24)
* odl_meetbot (3)
* LuisGomez (1)
* jamoluhrsen (1)



Generated by `MeetBot`_ 0.1.4