Comment 22 for bug 1518501
Revision history for this message
| OpenContrail Admin (ci-admin-f) wrote A change has been merged | #22 |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
bbfa235
(Get the code!)
Reviewed: https:/
Committed: http://
Submitter: Zuul
Branch: master
commit a56f464f0724811
Author: Manish <email address hidden>
Date: Tue Dec 8 22:59:03 2015 +0530
Mpls label from index vector not reused.
Problem:
In multicast evpn label, allocation happens from regular unicast range. When
this label is deleted it is not reset in index vector(Index vector manages
labels).
This used to result in freed label not being re-used by new allocations.
Solution:
Multicast has two sets of labels. First is fabric replication labels which are
reserved at init itself and remains till end of life.
Second set ie evpn label as mentioned above is allocated from non reserved
range. Both these labels are of type MCAST_NH. Now when evpn label is freed the
destructor used to check for MCAST_NH and skip de-allocation from index vector.
Fix is to check if mcast label is from fabric label reserved range and then
refrain from de-allocation. In all other case do the de-allocation.
Closes-bug: 1518501
Conflicts:
src/vnsw/
Conflicts:
src/vnsw/
src/vnsw/
Change-Id: Idd584a84d30f76
(cherry picked from commit 6389d9cd0106b8f
Agent crash @Filltrace, AgentRouteTable
Problem:
DeleteAllBgpPath deletes all path from CN. While deleting there may be some
paths which internally delete dependant paths and hence invalidating path
iterator.
Similar issue was observed here in bridge_route.cc:
https:/
Solution:
Maintain a list of paths to be deleted and once path list is traversed,
go through this list and delete paths.
Change-Id: I72eedd275ddf8d
Closes-bug: 1524140
(cherry picked from commit 7e51c5e5aa75348
Agent crash@ invalid path
Problem:
Deleteallbgppee
deleted. This used to include stale and non-stale paths from CN.
Later it will run over the list and delete them. During deletion of path by
fabric_
This is because it assumes that any non bgp peer path delete means route is no
more valid and no need of stale paths.
Now after deleting fabric_
remainders, it used to find stale path pointer(which as explained above gets
deleted by
fabric_
Solution:
Dont push stale paths in to_be_deleted_paths list.
Once all paths are deleted from to_be_deleted_
of stale paths, if any left.
Basically isolate deletion of stale and non stael paths.
Change-Id: I2ae02507750271
Closes-bug: 1524140
(cherry picked from commit e8dfd5b94f3ef1c
Conflicts:
src/vnsw/
Agent crashed because of freed xmpp channel.
Problem:
When agent xmpp channel is deleted, event to bring down bgp peer associated with
same is issued. This in turn starts a walk to clear all states and path
associated with this bgp peer. On finishing walk it unregisters all table
listeners. Meanwhile there can be a notification to controller on deleted peer.
This notification uses channel to verify if peer is deleted and ignore
notification. In this case channel was also deleted and hence crash happens.
Solution:
Defer deletion of channel till bgp peer is deleted.
This will not be executed if channel is not deleted but keeps flapping.
Closes-bug: 1524136
Change-Id: Ia43cbda1d9d2e8
(cherry picked from commit 887f5a2572767d8
Agent crash@ MplsTable::Add
Problem:
Crash was seen at creation of mpls label, where label itself was de-allocated
and re-allocation of same was not done.
This happened for multicast route where same label(evpn) is used across multiple
paths in route - local path, BGP peer(every). If BGP peer path goes away
recompute of master path for multicast used to by default delete label. In same
routine it used to find any other path having evpn label. If any is seen then
creation for same label was enqueued.
After this operation in DB for same label there is a delete and add operation.
Because of yielding policy delete may get executed and de-alloc label. Now when
add is executed agent crashes as label is not allocated.
If both these request get executed at once then this crash will not be seen.
Solution:
Delete evpn label only when non bgp peer path i.e. local vm peer goes off.
Any path with BGP peer will be using agent allocated label.
Along with this make all calls to add/delete mpls via recompute path of
multicast as inline.
Change-Id: I96030915be00ac
Closes-bug: 1525737
(cherry picked from commit 3dcc27a25357cca
Stale label in TSN.
Problem:
In RecomputeMultic
local_vm_peer path was not present. This was happening as it was populated from
other paths (tor/evpn) even in absence of local_vm_peer_path. This results in
stale label.
Example: No local vm peer path is present, tor path is present for flood route.
Now fabric path was added. RecomputeMultic
to presence of tor path. Addition of label has a check to add only if evpn label
is not zero and path peer is not BGP. In this case label was populated from tor
and peer is fabric_peer. Hence a stale label is added.
Solution:
Add label only when local vm peer path is added.
Change-Id: Ibc7c89b890ee7e
Closes-bug: 1525737
(cherry picked from commit 57728bb04fd1cd8
Conflicts:
src/vnsw/
Agent xmpp channel deleted.
Problem:
Statement is same as in- https:/
This didnt fix the issue completely. There was one case observed where two
requests were enqueued for bgp peer down on same channel. One got processed and
deleted channel before second one was dequeued. In this case channel should have
been retained for second response.
Solution:
Instead of keeping the mentioned fix, use shared_ptr for lifetime of channel.
All BGP peer take reference and agent xmpp channel in agent(class) take
reference. Once all are freed channel gets deleted.
Closes-bug: 1524136
Conflicts:
src/vnsw/
Change-Id: Id0cdbf737033a4
(cherry picked from commit a518cb88f8b15c9