The netlink deserialize is fetching information from the link.
This require the go routine to be in the correct namespace to
succeed
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Swarm mode does not really have anymore a use for the watchMiss.
Peer entries are configured at configuration time.
If the gcthresh denies the insertion the peerAdd will fail.
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
In case the file descriptor of the netlink socket is closed
the recvfrom is not returning. This may create deadlock conditions.
The current solution is to make sure that all the netlink socket used
have a proper timeout set on them to have the possibility to return
Added test to emulate the watchMiss condition
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Avoid error logs in case of local peer case, there is no need for deleteNeighbor
Avoid the network leave to readvertise already deleted entries to upper layer
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
In case of IP reuse locally there was a race condition
that was leaving the overlay namespace with wrong configuration
causing connectivity issues.
This commit introduces the use of setMatrix to handle the transient
state and make sure that the proper configuration is maintained
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
The package updated and now shows new warnings that had to be corrected
to let the CI pass
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Prevents an issue where the goroutine may jump to a new OS thread during
execution putting it into a mount/network NS that is unexpected.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
(cherry picked from commit 6d8617d8757a759d806a3307ca04d4d588c04aed)
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
In the peerDelete the updateDB flag was always true
In the peerAdd the updateDB flag was always true except for
the initSandbox case. But now the initSandbox is handled by the
go routing of the peer operations, so we can move that flag
down and remove it from the top level functions
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
The peerDbDelete was passing the wrong field to the underlay
Delete operation causing the mac entry to not being deleted
from the bridge on the overlay. This caused connectivity issue
when a container that before was remote was now scheduled
on the local node. The entry was such:
bridge fdb show | grep -i 02:42:0a:01:00:02
02:42:0a:01:00:02 dev vxlan0 master br0
02:42:0a:01:00:02 dev vxlan0 dst 172.31.14.63 link-netnsid 0 self permanent
That was still pointing to a remove node
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Move the sandbox init logic into the go routine that handles
peer operations.
This is to avoid deadlocks in the use of the pMap.Lock for the
network
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Remove the need for the wait group and avoid new
locks
Added utility to print the method name and the caller name
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
Remove the need for the wait group and avoid new
locks
Added utility to print the method name and the caller name
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
neighbor entries. On an l3 miss try to reprogram the neighbor entry
if the peer is valid. Its a best effort attempt because if the arp
table is still at gc_thresh3 value, addition will fail.
Signed-off-by: Santhosh Manohar <santhosh@docker.com>
The feature was not getting properly triggered, move it as
first operation in the configure
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
The netlink socket that was used to monitor the L2
miss was never being closed. The watchMiss goroutine
spawned was never returning. This was causing goroutine
leak in case of createNetwork/destroyNetwork
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
On linux systems bump up gc_thresholds so to lower the
probability of running with neighbor table overflow issues
Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>
- It specifies whether the network driver can
provide containers connectivity across hosts.
- As of now, the data scope of the driver was
being overloaded with this notion.
- The driver scope information is still valid
and it defines whether the data allocation
of the network resources can be done globally
or only locally.
- With the scope network option, user can now
force a network as swarm scoped
regardless of the driver data scope.
- In case the network is configured as swarm scoped,
and the network driver is multihost capable,
a network DB instance will be launched for it.
Signed-off-by: Alessandro Boch <aboch@docker.com>
Use the string concatenation operator instead of using Sprintf for
simple string concatenation. This is usually easier to read, and allows
the compiler to detect problems with the type or number of operands,
which would be runtime errors with Sprintf.
Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
