Reuse Cassandra connections

[Petr Malik]

Cassandra currently creates a new connection for every admin action and heartbeat check.

Connections are closed but there is apparently a resource leak in the engine causing the instance to run out of memory after some time resulting in the server getting killed by the kernel.

The guestagent should cache and reuse the connections in the Admin and Status objects.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/304316

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/mitaka
Review: https://review.openstack.org/313606

[OpenStack Infra (hudson-openstack)]

Change abandoned by amrith (<email address hidden>) on branch: stable/mitaka
Review: https://review.openstack.org/313606
Reason: need to cherry pick again

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/304316
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=728df1640f0db27fc541ad368341c98c1e77ef2c
Submitter: Jenkins
Branch: master

commit 728df1640f0db27fc541ad368341c98c1e77ef2c
Author: Petr Malik <email address hidden>
Date: Tue Apr 5 15:52:15 2016 -0400

    Reuse Cassandra connections

    Cache and reuse the connections in the Admin and
    Status objects to avoid leaking resources.
    Initialize the App and Admin objects themselves lazily
    at the first access.

    The Connection object has been changed to act as both
    an instance object and context manager.

    Change-Id: I4263b870a1099d345ab1940c61ccdf4680a706ca
    Closes-Bug: 1566946

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/313606
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=4175cb351fd8e3c9099fe592a434cd91dc847b1d
Submitter: Jenkins
Branch: stable/mitaka

commit 4175cb351fd8e3c9099fe592a434cd91dc847b1d
Author: Petr Malik <email address hidden>
Date: Tue Apr 5 15:52:15 2016 -0400

    Reuse Cassandra connections

    Cache and reuse the connections in the Admin and
    Status objects to avoid leaking resources.
    Initialize the App and Admin objects themselves lazily
    at the first access.

    The Connection object has been changed to act as both
    an instance object and context manager.

    Change-Id: I4263b870a1099d345ab1940c61ccdf4680a706ca
    Closes-Bug: 1566946
    (cherry picked from commit 728df1640f0db27fc541ad368341c98c1e77ef2c)

[Doug Hellmann (doug-hellmann)]

This issue was fixed in the openstack/trove 5.0.1 release.

[Doug Hellmann (doug-hellmann)]

This issue was fixed in the openstack/trove 6.0.0.0b2 development milestone.