heat engine restarts after some interval

[Tanvir]

Upon monitoring the heat-engine, it keeps on restarting after a particular interval.

contrail-version:

contrail-install-packages 2.23-129~juno 129

[Tanvir (tanvir-kekan)]

In /etc/coontrail/contrail-topology.conf the analytics api needs to point to the VIP in HA setup.

[Praneet Bachheti (praneetb)]

In an HA setup the /opt/contrail/bin/contrail-cmon-monitor.sh is restarting haproxy(on non-HA node) constantly as it sees clients connected to it. Due to this heat-engine fails to connect to haproxy (heat-api port) and exits.

As Sanju suggested .. "In /etc/coontrail/contrail-topology.conf the analytics api needs to point to the VIP in HA setup"

I will assign this bug to Sanju.

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20825
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20826
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20827
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20828
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20825
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20826
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20827
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20828
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/20828
Committed: http://github.org/Juniper/contrail-provisioning/commit/4524220f82cb25e13f6213bdc110f399521cc3b3
Submitter: Zuul
Branch: R3.0

commit 4524220f82cb25e13f6213bdc110f399521cc3b3
Author: Sanju Abraham <email address hidden>
Date: Wed Jun 1 00:02:37 2016 -0700

All client connecting to analytics api need to use the VIP if deployed in a HA setup.
The VIP on the LB node is what should recevive requests and proxy those to the backends.

Since cluster monitor will ensure that all TCP connections are on the active LB especially
after failures, any stale connections on the standby LB is forcefully reset.

With Anycast VIP, the reset will not happen as each LB instance will be able to proxy requests
to backends.

Closes Bug: #1578847

Change-Id: I3087893b0501d41cedf5eb89a6b71f7e38f5a0d0

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20922
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20825
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20826
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Review in progress for https://review.opencontrail.org/20827
Submitter: Sanju (<email address hidden>)

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/20922
Committed: http://github.org/Juniper/contrail-provisioning/commit/64265aad5206c0292542a44b9770950d6a0b4050
Submitter: Zuul
Branch: R3.0

commit 64265aad5206c0292542a44b9770950d6a0b4050
Author: Sanju Abraham <email address hidden>
Date: Fri Jun 3 08:26:41 2016 -0700

Closes Bug: #1578847. Typo corrected

Change-Id: Ic49dc4bccfed5b89892339b1d852757aee02d90e

[OpenContrail Admin (ci-admin-f)]

Reviewed: https://review.opencontrail.org/20825
Committed: http://github.org/Juniper/contrail-provisioning/commit/3f2029f3ea72e46d886f0f2c2660f9f26aeee44d
Submitter: Zuul
Branch: master

commit 3f2029f3ea72e46d886f0f2c2660f9f26aeee44d
Author: Sanju Abraham <email address hidden>
Date: Tue May 31 23:48:02 2016 -0700

All client connecting to analytics api need to use the VIP if deployed in a HA setup.
The VIP on the LB node is what should recevive requests and proxy those to the backends.

Since cluster monitor will ensure that all TCP connections are on the active LB especially
after failures, any stale connections on the standby LB is forcefully reset.

With Anycast VIP, the reset will not happen as each LB instance will be able to proxy requests
to backends.

Closes Bug: #1578847

Change-Id: If2b41876dcff976e14e3fa87dcb7840be928cec4