Age | Commit message (Collapse) | Author | Files | Lines |
|
Previously we were always run the galera-ready exec every step. This
change switches it to be refreshonly so we only wait when the service is
setup or restarted.
Change-Id: I5ff9d49c2590751913b96777bcd72c8a15627a01
Closes-Bug: #1680586
|
|
|
|
|
|
This reverts commit 3f7e74ab24bb43f9ad7e24e0efd4206ac6a3dd4e.
After identifying how to workaround the performance issues on the
undercloud, let's put this back in. Enabling innodb_file_per_table is
important for operators to be able to better manage their databases.
Change-Id: I435de381a0f0e3ef221e498f442335cdce3fb818
Depends-On: I77507c638237072e38d9888aff3da884aeff0b59
Closes-Bug: #1660722
|
|
This reverts commit 621ea892a299d2029348db2b56fea1338bd41c48.
We're getting performance problems on SATA disks.
Change-Id: I30312fd5ca3405694d57e6a4ff98b490de388b92
Closes-Bug: #1661396
Related-Bug: #1660722
|
|
InnoDB uses a single file by default which can grow to be
tens/hundreds of gigabytes, and is not shrinkable even
if data is deleted from the database.
Best practices are that innodb_file_per_table is set to ON
which instead stores each database table in its own file, each of
which is also shrinkable by the InnoDB engine.
Closes-Bug: #1660722
Change-Id: I59ee53f6462a2eeddad72b1d75c77a69322d5de4
|
|
Follow up patch for I63da4f48da14534fd76265764569e76300534472
to support composable HA for the Ceph rbdmirror daemon.
Change-Id: I3767bee4b1c7849fa85e71bcc57534b393d2d415
|
|
|
|
|
|
This commit implements composable HA for the pacemaker profiles.
- Everytime a pacemaker resource gets included on a node,
that node will add a node cluster property with the name of the resource
(e.g. galera-role=true)
- Add a location rule constraint to force running the resource only
on the nodes that have that property
- We also make sure that any pacemaker resource/property creation has a
predefined number of tries (20 by default). The reason for this is
that within composable HA, it might be possible to get "older CIB"
errors when another node changed the CIB while we were doing an
operation on it. Simply retrying fixes this.
- Also make sure that we use the newly introduced
pacemaker::constraint::order class instead of the older
pacemaker::constraint::base class. The former uses the push_cib()
function and hence behaves correctly in case multiple nodes try
to modify the CIB at the same time.
Change-Id: I63da4f48da14534fd76265764569e76300534472
Depends-On: Ib931adaff43dbc16220a90fb509845178d696402
Depends-On: I8d78cc1b14f0e18e034b979a826bf3cdb0878bae
Depends-On: Iba1017c33b1cd4d56a3ee8824d851b38cfdbc2d3
|
|
Previously we missed to perform the basic Ceph client configuration
on a node where only the RBD mirror service was deployed.
Change-Id: Ie6a4284a88714bcee964a38636e12aa88bb95c9d
Co-Authored-By: Michele Baldessari <michele@acksyn.org>
Related-Bug: #1652177
|
|
There is a typo in the bootstrap check which will lead to:
Could not find data item ceph_rbdmirror_bootstrap_short_node_name in any
Hiera data file and no default supplied at
/etc/puppet/modules/tripleo/manifests/profile/pacemaker/ceph/rbdmirror.pp
We need to be using the correct one:
$ hiera ceph_rbdmirror_short_bootstrap_node_name
overcloud-remote-0
Change-Id: Ic343e5f99e48360bdd2d2989781a4b6ca484e8fc
|
|
This change adds a profile for the Ceph RBD mirror service, which
should be managed by Pacemaker to make sure there is always a single
instance running.
Change-Id: Ic63dc5cffece38942d305f538f71dd58a5d50789
Partial-Bug: #1652177
|
|
When we create a pacemaker resource it must happen from a single node.
If it happens from multiple nodes an immediate error will be returned by
pcs.
For the pacemaker roles we enforce this by leveraging the recently
introduced <SERVICE_NAME_bootstrap_short_node_name> which gives us
the first hostname per-service, regardless of the role.
(introduced via I03e8685f939e8ae1fcd8b16883b559615042505d)
With this approach if a pacemaker service belongs to two different
roles (say role Controller on node A and role galera on node B), it
will only create the resource from one of the two and not both (which
would return an error).
Only setting Partial-Bug for this one, because it addresses the issue
from the pacemaker resource creation POV (which is always affected). But
the issue itself is a race that we're theoretically affected by since
the composable roles work landed. While I have tried to fix the more
general case in previous attempts, I think it is best if we start a
discussion on how to fix it, because each approach has a bunch of
potential drawbacks and is quite invasive on how we do things. A
discussion slot for this has been proposed for the Atlanta PTG.
Change-Id: I662398cab60d523d204b57a5674ca8f5c0f2e68a
Partial-Bug: #1615983
|
|
Manila ceph driver reads ceph's client configuration
(keyring is the most important) from ceph.conf file
(or any other file set by cephfs_conf_path). ceph.conf
should be updated with keyring location.
If ceph is deployed by tripleo then also manila ceph key
is added into ceph and ceph filesystem is created.
Depends-On: I18436a64fc991b9e697a1d79e369ac110cf8fe20
Change-Id: Iac4a260af6738ed6afd4bcb107221a736d07c1b5
Partial-Bug: #1644784
Closes-Bug: #1646147
|
|
Removed redundant 'the'
Change-Id: Ie2051f35ec1e7010423c46084f5512c02af85f33
|
|
Manila pacemaker manifest (which sets manila share service only)
includes also manila api and scheduler manifests. There is no
reason for this. Also it causes that on whichever node manila share
service runs also manila api and scheduler services are started.
Change-Id: Ia1b39ef36c5bc34813cd6430b69ad9b698acc3cf
Closes-Bug: #1653500
|
|
By default galera-monitor xinetd is binding on all the interfaces.
That means that the port 9200 is exposed on the external network.
Because haproxy is using the same network for the backend and the
check we can reuse it for the xinetd binding.
Change-Id: If1a50515593e81f46d67309bdeecbe84c1d0ebe4
|
|
|
|
Puppet 4 ordering make things more strict in catalog, which is good.
Resources have to be orchestrated or Puppet will take them in the order
they are found in catalog.
This patch makes sure we create MySQL users only when Galera is actually
ready.
Closes-Bug: #1645787
Change-Id: I536a1a128c3a7eca49bcc4f34a1307bcd60b029e
|
|
|
|
If we upgrade a cloud that was configured with external load balancer
the process will fail during convergence step because it will try to
restart haproxy which is not configured when an external load balancer
is configured.
Closes-Bug: #1636527
Change-Id: I6f6caec3e5c96e77437c1c83e625f39649a66c48
|
|
With the landing of HA NG in Newton we can actually remove the
pacemaker profiles we do not need. The only ones that are being
used in one form or the other are:
$ grep -ir services\/pacemaker environments | awk '{ print $3 }' | sort | uniq
../puppet/services/pacemaker/cinder-backup.yaml
../puppet/services/pacemaker/cinder-volume.yaml
../puppet/services/pacemaker/database/mysql.yaml
../puppet/services/pacemaker/database/redis.yaml
../puppet/services/pacemaker/haproxy.yaml
../puppet/services/pacemaker/manila-share.yaml
../puppet/services/pacemaker/rabbitmq.yaml
../puppet/services/pacemaker.yaml
The only exception is profile/pacemaker/database/mongodbvalidator
because it is included by profile/base/database/mongodb.pp
Change-Id: I80c8559bb2d915385bcc20ae71fe144ddd6591c1
|
|
The current redis file descriptor limit is 4096 because of two reasons:
- It is run via the redis user
- It is not started via systemd which has explicit LimitNOFILE set to
10240 (which matches the default configuration of maximum 10000
clients)
Create an /etc/security/limits.d/redis.conf file in order to increase
the fd limit value With this change we correctly get the following
limits:
[root@overcloud-controller-0 ~]# pcs status |grep -A2 redis
Master/Slave Set: redis-master [redis]
Masters: [ overcloud-controller-2 ]
Slaves: [ overcloud-controller-0 overcloud-controller-1 ]
[root@overcloud-controller-0 ~]# cat /proc/`pgrep redis`/limits | grep open
Max open files 10240 10240 files
Previously this limit was set to 4096.
Change-Id: I7691581bad92ad9442cecd82cf44f5ac78ed169f
Closes-Bug: #1635334
|
|
|
|
remove_default_accounts is a mysql::server parameter that, set to True,
will execute some MySQL commands to cleanup MySQL defaults accounts
created by packaging.
In order to successfully run the commands, we need MySQL up and running,
which is not the case at step 1 but at step 2.
This patch make sure we run the commands at step 2 on pacemaker master
only.
No change for scenarios without Pacemaker.
Change-Id: Ifad3cb40fd958d7ea606b9cd2ba4c8ec22a8e94e
Closes-Bug: #1633113
|
|
When we observe the 'stop timeout' values of pacemaker resources:
rabbitmq and redis, they are set to 90s. But for all other services, it
is set to 200s.
The overcloud deployment sometimes fails due to this with the error:
Error: Could not complete shutdown of rabbitmq-clone, 1 resources
remaining
Error performing operation: Timer expired
This patch updates the timeout for Redis and RabbitMQ to avoid this
error.
Change-Id: I8a3b3951a896ee3e8e5e09778e8ea4717e76a1b4
|
|
Aligns the way how we check for enabled backends in
pacemaker/manila.pp with what we did in base/manila/api.pp with [1].
The benefit is that we don't need to emit from the templates
custom hiera.
1. I86ba8b9d5872c0f1a94e74215e97b796ad129bfb
Change-Id: I04e28a95e8d69a24cd3df109bf1802bfcbd941db
|
|
|
|
We're not able to use FQDNs yet, so to work around this, we give
precedence to a "short name" list we'll get from t-h-t.
Change-Id: I4ef7786474c229d5212a0deb2ca02ee992b030d8
Related-Bug: #1628521
|
|
It turns out that reducing number of rabbitmq queues in cluster
significantly improves performance of cluster especially in the case of
failover recovery time. Right now the cluster uses ha-all mode for rabbitmq
queues.
It is best to change this to "ha-exactly" mode and reduce the number
of queue copies to ceil(N/2) where N is number of controllers in the
cluster - so in typical scenario of 3 controller It would be 2 by
default.
It does not make much sense to keep the copies of queues over whole
cluster since if the quorum of nodes is lost then the rest of cluster
nodes will be stopped anyway. We let the user override this with a
parameter.
I.e. for a 3 node controlplane cluster we will go from this:
pcs resource show rabbitmq
Resource: rabbitmq (class=ocf provider=heartbeat type=rabbitmq-cluster)
Attributes: set_policy="ha-all ^(?!amq\.).* {"ha-mode":"all"}"
To this:
pcs resource show rabbitmq
Resource: rabbitmq (class=ocf provider=heartbeat type=rabbitmq-cluster)
Attributes: set_policy="ha-all ^(?!amq\.).* {"ha-mode":"exactly","ha-params":2}"
According to Marin Krcmarik's testing recovery time from failure was
reduced significantly.
Co-Authored-By: Marian Krcmarik <mkrcmari@redhat.com>
Change-Id: Ib62001c03e1e08f58cf0c6e0ba07a8879a584084
Partial-Bug: #1628998
|
|
Back in the Mitaka cycle via the change If6b43982c958f63bc78ad997400bf1279c23df7e
we made sure that the default start and stop timeouts for pacemaker
systemd resources is 200s (>= twice the default 90s DefaultTimeoutStopSec
in systemd). We did this change by setting puppet resource defaults for
the Pacemaker::Resource::Service class:
Pacemaker::Resource::Service {
op_params => 'start timeout=200s stop timeout=200s',
}
The problem is that after the composable services rework, this does not
work anymore and the pacemaker systemd resources that still exist do not
have these timeouts set.
We want to move away from resource defaults for this because its results
are dependent on the inclusion order which in tripleo is not guaranteed
any longer (https://docs.puppet.com/puppet/latest/reference/lang_scope.html#scope-lookup-rules)
The only services affected in Newton are: cinder-volume,
cinder-backup, manila-share, haproxy. I preferred fixing all the
pacemaker resources because it seems the cleanest and most logical
commit.
Change-Id: If89a95706514e536a7a2949871a0002c79b6046e
Closes-Bug: #1629366
|
|
|
|
This patch moves the various DB syncs into the MySQL role.
Database creation needs to occur on the MySQL server to
avoid permission issues.
This patch also moves database creation to step 2 so we can
guarantee that all per-service databases exist at this time.
This avoids complex ordering needed during step 3 where
services, on different hosts, can run their own db sync's
in a distributed fashion.
Change-Id: I05cc0afa9373429a3197c194c3e8f784ae96de5f
Partial-bug: #1620595
|
|
|
|
|
|
having an actual name for that configuration will allow us to pass a
more proper name via t-h-t.
Change-Id: Iea4bd67074824e5dc6732fd7e408743e693d80b3
|
|
It used to be hardcoded that the bind-address was always coming from
the $::hostname fact. This is wrong, as it disregards where we have
configured the mysql address. This commit actually makes it
configurable, so we'll be able to set it via hieradata.
On the other hand, we use the hiera key that we already set
'mysql_bind_host' as a default; if, for some reason, that's
unavailable then we fall back to $::hostname.
Related-Bug: #1627060
Change-Id: I316acfd514aac63b84890e20283c4ca611ccde8b
|
|
In puppet-manila it is the api service performing db sync, not
scheduler. This change moves ::manila::db::mysql (which creates
the empty database and users) in the tripleo manila/api profile.
Also moves rabbit config into a general manila base profile as
that would be needed by the scheduler service as well.
Change-Id: I2b537f735b8d1be8f39e8c274be3872b193c1014
|
|
The puppet-tripleo side for manila-cephfs landed without specifying
defaults for all class params [1] so when cephfs isn't enabled
e.g. only generic, then you will get errors for those params. See
review comments at [2] for reports of this.
This will fixup the manila-cephfs puppet-tripleo side to be more
in line with the tidy up adding netapp at [3]. The config is all
moved tripleo-heat-templates side. The tht review for this
is at https://review.openstack.org/#/c/358525/ and that will now
depend on this review.
[1] https://review.openstack.org/#/c/354047/
[2] https://review.openstack.org/#/c/354019/
[3] https://review.openstack.org/#/c/354014/
Change-Id: I918f6f23ae0bd3542bcfe1bf0c797d4e6aa8f4d9
|
|
|
|
This adds support for the manila-netapp backend. The backend
specific config is set tht side. So this change also
tidies up the manila generic config, which is unnecessarily
being duplicated here
( see https://review.openstack.org/#/c/354019/ )
Change-Id: Ic6f8e8d27ca20b9badddea5d16550aa18bff8418
|
|
Don't add brackets on mysql_bind_host parameter in Galera config.
Having brackets from this parameter works with old version of
Galera but not newest one.
So let's remove them at all, so we can safely upgrade Galera in RDO.
Change-Id: Ic904d4efda162f18ec8dffb91c2f383f54361f41
Closes-Bug: #1622755
|
|
|
|
|
|
Write restart flag file for services managed by Pacemaker into
/var/lib/tripleo/pacemaker-restarts directory. The name of the file must
match the name of the clone resource defined in pacemaker. The
post-puppet restart script will restart each service having a restart
flag file and remove those files.
This approach focuses on $pacemaker_master only (we don't want to
restart the pacemaker services 3 times when we have 3 controllers), so
it relies on the assumption that we're making the matching config
changes across the pacemaker nodes.
Change-Id: I6369ab0c82dbf3c8f21043f8aa9ab810744ddc12
|
|
profiles"
|
|
Prepare the pacemaker mysql manifest that galera_node_names will be an
array. Stay backwards compatible to handle comma-delimited-string too
and avoid a chicken-and-egg patch problem between t-h-t and
puppet-tripleo.
Change-Id: Ia0d9d59728c8771974bfbc486f4929b99a38e4fb
Partially-Implements: blueprint custom-roles
|
|
|
|
Change-Id: Idaad75238a2884fe82b1e5fce3ed910d866b98a2
|