New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CLEARWATER: CA-96541/SCTX-1188: Fix race condition when creating metrics for VLAN PI... #1084

Merged
merged 1 commit into from Apr 6, 2013

Conversation

Projects
None yet
2 participants
@robhoes
Member

robhoes commented Mar 15, 2013

...Fs

This fixes a race condition between the thread that copies bonds and VLANs from
the master when a new slave joins the pool, and the thread that creates/updates
the PIF_metrics (the "monitor thread"). Or in fact a race between anything that
creates a bond and very soon after that a VLAN on top of it.

Since XCP1.6/XS6.1, VLANs no longer have their own PIF_metrics. Rather, they point
to the PIF_metrics of the underlying PIF, which can be either a physical PIF
(e.g. for eth0), or a bond PIF. Therefore, the carrier of a VLAN PIF is the same
as the carrier of the underlying PIF.

When a physical PIF or bond PIF is created, it does not immediately have
PIF_metrics (PIF.metrics = "OpaqueRef:NULL"). The PIF_metrics object is created
and assigned to the PIF by the monitor thread a short while after that. If you
create a VLAN before the PIF_metrics have been created, the PIF.metrics field from
the underlying PIF will still be copied, which is "OpaqueRef:NULL". This will never
get updated, because the monitor thread ignores VLANs.

The fix is simply to immediately create an (empty) metrics object when creating a
bond.

Duplicate of #1083

CA-96541/SCTX-1188: Fix race condition when creating metrics for VLAN…
… PIFs

This fixes a race condition between the thread that copies bonds and VLANs from
the master when a new slave joins the pool, and the thread that creates/updates
the PIF_metrics (the "monitor thread"). Or in fact a race between anything that
creates a bond and very soon after that a VLAN on top of it.

Since XCP1.6/XS6.1, VLANs no longer have their own PIF_metrics. Rather, they point
to the PIF_metrics of the underlying PIF, which can be either a physical PIF
(e.g. for eth0), or a bond PIF. Therefore, the carrier of a VLAN PIF is the same
as the carrier of the underlying PIF.

When a physical PIF or bond PIF is created, it does not immediately have
PIF_metrics (PIF.metrics = "OpaqueRef:NULL"). The PIF_metrics object is created
and assigned to the PIF by the monitor thread a short while after that. If you
create a VLAN before the PIF_metrics have been created, the PIF.metrics field from
the underlying PIF will still be copied, which is "OpaqueRef:NULL". This will never
get updated, because the monitor thread ignores VLANs.

The fix is simply to immediately create an (empty) metrics object when creating a
bond.

Signed-off-by: Rob Hoes <rob.hoes@citrix.com>

@ghost ghost assigned malcolmcrossley Mar 20, 2013

malcolmcrossley pushed a commit that referenced this pull request Apr 6, 2013

Malcolm Crossley
Merge pull request #1084 from robhoes/ca96541-cw
CLEARWATER: CA-96541/SCTX-1188: Fix race condition when creating metrics for VLAN PI...

@malcolmcrossley malcolmcrossley merged commit baf28d0 into xapi-project:clearwater Apr 6, 2013

1 check was pending

default Merged build triggered.

@robhoes robhoes deleted the robhoes:ca96541-cw branch Apr 12, 2013

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment