Optimization: Reduce the number of transactions to OVSDBs

[Lucas Alvares Gomes]

After the patch to check for the agents health status was introduced, networking-ovn started to bump the nb_cfg in NBDB for every transaction.

That behavior ended up increasing the number of transactions because every time nb_cfg is increased that would be propagated to the SBDB and the metadata agent would get an event which would result in it starting yet another transaction to update its Chassis external_ids column (SBDB) to indicate that it's still alive.

To reproduce (on DevStack):

1. Listen to the events to the SBDB

$ tailf /opt/stack/data/ovs/ovnsb_db.db

2. Start a transaction, in this example I will clean the port_bind table which will trigger the port to go down and up again:

$ ovn-sbctl clear port_bind ae2db6eb-26d1-401e-aa0f-c866ec9974f3 chassis

3. Now looking at the transactions we will see that we have 8 operations in the SBDB:

OVSDB JSON 198 979a9514e40bec7e4fb21c134b4d2d417b1463b3
{"_date":1549290530763,"Port_Binding":{"96837e20-2535-4e3e-8361-49e25107adaf":{"chassis":["set",[]]}},"_comment":"ovs-sbctl: ovn-sbctl clear port_bind 96837e20-2535-4e3e-8361-49e25107adaf chassis"}
OVSDB JSON 140 ecb2b62ffd3c83ecbd1ce5bc914f3678b7968b81
{"_date":1549290530769,"Port_Binding":{"96837e20-2535-4e3e-8361-49e25107adaf":{"chassis":["uuid","d3cbc656-2e88-4e51-a412-44e2aa327c1a"]}}}
OVSDB JSON 93 b1675d63e5f4ff2ce24a33247dafe9bd345e8537
{"SB_Global":{"21cbcd3d-2dca-47e5-b964-b4195aabee75":{"nb_cfg":9593}},"_date":1549290531413}
OVSDB JSON 91 b2b716ba0b16ae2e4a4f86cc2404022e39469c5d
{"Chassis":{"d3cbc656-2e88-4e51-a412-44e2aa327c1a":{"nb_cfg":9593}},"_date":1549290531421}
OVSDB JSON 405 0a7ea709527baa2a2defe1a9e1adbd6c719eb68c
{"Chassis":{"d3cbc656-2e88-4e51-a412-44e2aa327c1a":{"external_ids":["map",[["datapath-type",""],["iface-types","erspan,geneve,gre,internal,ip6erspan,ip6gre,lisp,patch,stt,system,tap,vxlan"],["neutron:ovn-metadata-id","f6c9cc58-47c7-49b9-bd6e-94f7aba12e9d"],["neutron:ovn-metadata-sb-cfg","9593"],["ovn-bridge-mappings","public:br-ex"],["ovn-cms-options","enable-chassis-as-gw"]]]}},"_date":1549290531428}
OVSDB JSON 531 35e95217c6ede96f56602ef18d19fb9f3ea1fe56
{"_date":1549290531579,"SB_Global":{"21cbcd3d-2dca-47e5-b964-b4195aabee75":{"nb_cfg":9594}},"Port_Binding":{"dafe6ad6-98b4-4691-b963-3c026d817367":{"external_ids":["map",[["neutron:cidrs","172.24.4.228/24 2001:db8::171/64"],["neutron:device_id","3e24f3fb-87c1-4e33-a132-f63ad76039f5"],["neutron:device_owner","network:router_gateway"],["neutron:network_name","neutron-25bbb5aa-7e59-415f-b12f-9ed7a6049a76"],["neutron:port_name",""],["neutron:project_id",""],["neutron:revision_number","210"],["neutron:security_group_ids",""]]]}}}
OVSDB JSON 91 dd6b725b6a875318d7207c4565d09f2acd66308e
{"Chassis":{"d3cbc656-2e88-4e51-a412-44e2aa327c1a":{"nb_cfg":9594}},"_date":1549290531586}
OVSDB JSON 405 139e6404aec4701d62757479ce1e8c7b940ead2c
{"Chassis":{"d3cbc656-2e88-4e51-a412-44e2aa327c1a":{"external_ids":["map",[["datapath-type",""],["iface-types","erspan,geneve,gre,internal,ip6erspan,ip6gre,lisp,patch,stt,system,tap,vxlan"],["neutron:ovn-metadata-id","f6c9cc58-47c7-49b9-bd6e-94f7aba12e9d"],["neutron:ovn-metadata-sb-cfg","9594"],["ovn-bridge-mappings","public:br-ex"],["ovn-cms-options","enable-chassis-as-gw"]]]}},"_date":1549290531596}

Now, when we disable the increase of nb_cfg (by commenting the code, there's no config for it) this number of transactions is reduced to 3 (same command above):

OVSDB JSON 198 c588ca88fbbcdbacbedbe6b6b1ddcf897f84606d
{"_date":1549289886733,"Port_Binding":{"96837e20-2535-4e3e-8361-49e25107adaf":{"chassis":["set",[]]}},"_comment":"ovs-sbctl: ovn-sbctl clear port_bind 96837e20-2535-4e3e-8361-49e25107adaf chassis"}
OVSDB JSON 140 26497f926d2b3c0f906f9fc0030cd56233b6014a
{"_date":1549289886740,"Port_Binding":{"96837e20-2535-4e3e-8361-49e25107adaf":{"chassis":["uuid","d3cbc656-2e88-4e51-a412-44e2aa327c1a"]}}}
OVSDB JSON 462 a69244cf4999af5fcda3145a95f830c6d17ffb12
{"_date":1549289887593,"Port_Binding":{"dafe6ad6-98b4-4691-b963-3c026d817367":{"external_ids":["map",[["neutron:cidrs","172.24.4.228/24 2001:db8::171/64"],["neutron:device_id","3e24f3fb-87c1-4e33-a132-f63ad76039f5"],["neutron:device_owner","network:router_gateway"],["neutron:network_name","neutron-25bbb5aa-7e59-415f-b12f-9ed7a6049a76"],["neutron:port_name",""],["neutron:project_id",""],["neutron:revision_number","208"],["neutron:security_group_ids",""]]]}}}

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/634912

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/634912
Committed: https://git.openstack.org/cgit/openstack/networking-ovn/commit/?id=9ee16bf283e612216a0be0a3d8f86e0263e0cc55
Submitter: Zuul
Branch: master

commit 9ee16bf283e612216a0be0a3d8f86e0263e0cc55
Author: Lucas Alvares Gomes <email address hidden>
Date: Mon Feb 4 16:21:23 2019 +0000

    Increase nb_cfg when needed only (reducing the # of transactions)

    Prior to this patch, any transaction to NBDB would result in the nb_cfg
    from NBDB being increased by 1. That would propagated to the SBDB and
    the metadata agent would get an event which would result in yet another
    transaction being created to update the Chassis' external_ids (SBDB)
    column to inform that the agent was alive.

    This patch is changing that behavior to bump the nb_cfg only when
    the get_agents() API call is invoked which would reduce the number
    of transactions created in the SBDB due to our current health check
    mechanism.

    Closes-Bug: #1814698
    Change-Id: I6f91955444e0601f10f2d5e509d4a4453def5f2b
    Signed-off-by: Lucas Alvares Gomes <email address hidden>

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/networking-ovn 6.0.0.0b1 development milestone.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/rocky
Review: https://review.openstack.org/648160

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/648160
Committed: https://git.openstack.org/cgit/openstack/networking-ovn/commit/?id=d90c7915d9cbd81e6b772f8ebbd2c8ca9cffa9a1
Submitter: Zuul
Branch: stable/rocky

commit d90c7915d9cbd81e6b772f8ebbd2c8ca9cffa9a1
Author: Lucas Alvares Gomes <email address hidden>
Date: Mon Feb 4 16:21:23 2019 +0000

    Increase nb_cfg when needed only (reducing the # of transactions)

    Prior to this patch, any transaction to NBDB would result in the nb_cfg
    from NBDB being increased by 1. That would propagated to the SBDB and
    the metadata agent would get an event which would result in yet another
    transaction being created to update the Chassis' external_ids (SBDB)
    column to inform that the agent was alive.

    This patch is changing that behavior to bump the nb_cfg only when
    the get_agents() API call is invoked which would reduce the number
    of transactions created in the SBDB due to our current health check
    mechanism.

    Closes-Bug: #1814698
    Change-Id: I6f91955444e0601f10f2d5e509d4a4453def5f2b
    Signed-off-by: Lucas Alvares Gomes <email address hidden>
    (cherry picked from commit 9ee16bf283e612216a0be0a3d8f86e0263e0cc55)

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/networking-ovn 5.1.0 release.