Threads are the recommended way to scale CPU performance since HAProxy
1.8.
Official documentation says: « While "nbproc" historically used to be the only
way to use multiple processors, it also involved a number of shortcomings
related to the lack of synchronization between processes (health-checks, peers,
stick-tables, stats, ...) which do not affect threads. As such, any modern
configuration is strongly encouraged to migrate away from "nbproc" to
"nbthread". ».
Change-Id: I6f2e9d74e68703c8e0827e495945a75f020e1561
The directive used has the same semantic as what is done above for nbproc > 1:
it binds each thread to a CPU. It is simpler and does not require a loop because
it uses the auto: syntax available in HAProxy 2.4.
Change-Id: I1ce124b678140f5f4737df557683bb67bc7cfc66
Threads are the recommended way to scale CPU performance since HAProxy
1.8.
Official documentation says: « While "nbproc" historically used to be the only
way to use multiple processors, it also involved a number of shortcomings
related to the lack of synchronization between processes (health-checks, peers,
stick-tables, stats, ...) which do not affect threads. As such, any modern
configuration is strongly encouraged to migrate away from "nbproc" to
"nbthread". ».
While more recent versions of HAProxy automatically detect the number of
available CPU and enable threads for them, it can be useful to explicitely set
the value.
In this patch, setting cpu-map for threads is not supported.
Change-Id: Id917c70f3dbe52f24f25d9403ba8151729e8966b
Use case: exposing single external https frontend and
load balancing services using FQDNs.
Support different ports for internal and external endpoints.
Introduced kolla_url filter to normalize urls like:
- https://magnum.external:443/v1
- http://magnum.external:80/v1
Change-Id: I9fb03fe1cebce5c7198d523e015280c69f139cd0
Co-Authored-By: Jakub Darmach <jakub@stackhpc.com>
We've seen issues in CI when keepalived haproxy check script returns
an error and keepalived is switching to backup and then again to primary
on a single node environment.
Closes-Bug: #2025219
Change-Id: Iba62e76b3cf83f3ade6df81288d2d77129ffc725
By default ProxySQL's default value of max_replication_lag
is 0 which is in fact disabling this feature [1].
If it is greater than 0, ProxySQL will regularly monitor
replication lag and if it goes beyond the configured threshold
it will temporary shun the host until replication catches up.
This should be configurable via kolla-ansible as every
openstack deployment can be different in terms of network
delays, database load etc.. , so user should have option
to configure when database backend will be shunned.
[1] https://proxysql.com/documentation/main-runtime/
Change-Id: I66171638abc712cb84b380042f1d29f54c499e73
Kolla environment currently uses haproxy
to fullfill HA in mariadb. This patch
is switching haproxy to proxysql if enabled.
This patch is also replacing mariadb's user
'haproxy' with user 'monitor'. This replacement
has two reasons:
- Use better name to "monitor" galera claster
as there are two services using this user
(HAProxy, ProxySQL)
- Set password for monitor user as it's
always better to use password then not use.
Previous haproxy user didn't use password
as it was historically not possible with
haproxy and mariadb-clustercheck wasn't
implemented.
Depends-On: https://review.opendev.org/c/openstack/kolla/+/769385
Depends-On: https://review.opendev.org/c/openstack/kolla/+/765781
Depends-On: https://review.opendev.org/c/openstack/kolla/+/850656
Change-Id: I0edae33d982c2e3f3b5f34b3d5ad07a431162844
Allow operators to set haproxy socket to admin level.
This is done via the flag haproxy_socket_level_admin which
is set to "no" by default.
Closes-Bug: 1960215
Signed-off-by: Imran Hussain <ih@imranh.co.uk>
Change-Id: Ia0da89288d68f5803ace1934c013053f12343195
The check condition is similar to the previous if and it can be merged
Signed-off-by: Seena Fallah <seenafallah@gmail.com>
Change-Id: I62213d91945de42ffc87ecad1e96fa4fc0760f10
For now role haproxy is maintaining haproxy
and keepalived. In follow-up changes there is also
proxysql added.
This patch is *only* renaming/moving stuff to more
prominent role loadbalancer, and moving also specific
templates to subdirectory.
This was done only to better diff in follow-up
changes.
Change-Id: I1d39d5bcaefc4016983bf267a2736b742cc3a555
