Commit Graph

115 Commits

Author SHA1 Message Date
Joshua Boniface 0ee5a93b2d Use debian_version custom fact 2023-08-08 16:27:51 -04:00
Joshua Boniface 7b51fe0d00 Adjust name of bootstrap trigger variable
The PVC bootstrap framework overrides this variable and wreaks havoc on
it. Instead adjust our side so that it looks for do_bootstrap instead.
2023-01-16 14:39:43 -05:00
Joshua Boniface 3c1c97dbf5 Ensure CPU tuning is only applied on Debian 11+ 2022-11-15 16:21:48 -05:00
Joshua Boniface c5ec0f4f62 Add updated tuning configuration
Uses a much nicer CPU tuning configuration, leveraging systemd's
AllowedCPUs and CPUAffinity options within a set of slices (some
default, some custom).

Configuration is also greatly simplified versus the previous
implementation, simply asking for a number of CPUS for both the system
and OSDs, and calculating everything else that is required.

Also switches (back) to the v2 unified cgroup hierarchy by default as
required by the systemd AllowedCPUs directive.
2022-11-13 01:33:09 -05:00
Joshua Boniface 7d329708bc Remove cpuset configurations
This functionality simply did not work, with Libvirt continuing to dump
its processes into the root cset thus defeating the purpose entirely.

Just remove it, from some very initial testing it isn't worth the
headache.
2022-11-06 01:08:27 -04:00
Joshua Boniface 0dbcf10d4a Add rinse dependency for provisioner 2022-10-07 00:41:03 -04:00
Joshua Boniface 503a2e6c0b Remove pvc-flush references
This service causes more problems than it solves usually, so it is being
removed in the next PVC version.
2022-07-25 23:19:38 -04:00
Joshua Boniface 87ae5a12e9 Add Ceph support for single-node clusters
Ensures that the pool default size/min size is set to something
reasonable for a single node (effective RAID-1) and replace teh default
CRUSH replicate_rule set for this situation with one choosing OSD
instead of host as the default.
2022-07-08 09:33:28 -04:00
Joshua Boniface a213ee7d4a Add cgroup delegation override
Required to solve the occasional
  libvirt: QEMU Driver error : Requested operation is not valid:
  cgroup CPUACCT controller is not mounted
problem, as per:
  https://answers.launchpad.net/ubuntu/+question/665132
2022-06-03 12:01:08 -04:00
Joshua Boniface dd4825a30a Ensure packages are installed as newhost 2022-04-26 11:10:11 -04:00
Joshua Boniface edcf14a78c Ensure Admin users are in additional groups 2022-01-04 15:13:13 -05:00
Joshua Boniface 7b07a81eca Convert default libvirtd to template 2022-01-01 01:50:54 -05:00
Joshua Boniface 4b6fdf301f Make locale generation universal
Don't rely on a notify/handler, just do it every time in the base role.
2021-12-28 14:57:25 -05:00
Joshua Boniface d859e032be Ensure insecure_global_id_reclaim is false 2021-12-28 02:06:38 -05:00
Joshua Boniface e48b6eef3a Fix bugs with Patroni bootstrap 2021-12-20 16:52:53 -05:00
Joshua Boniface 908982e97d Add proper PostgreSQL versioning 2021-12-15 12:11:49 -05:00
Joshua Boniface f8898f22b4 Ensure all zkCli has -server set 2021-12-15 11:29:10 -05:00
Joshua Boniface a9d636dcf5 Fix incorrect postgresql version 2021-12-15 02:22:10 -05:00
Joshua Boniface bea79b5102 Add immutability to PVC subrole
1. Remove the obsolete pvc-vacuum script install.

2. Remove notifies when modifying configs; we do not want to restart the
daemons uncontrolled.

3. Add bootstrap check to package installs so they only happen on
bootstrap.

This ensures this part of the role, on re-runs, will *only* update
configs and not actually touch the running daemon. This makes it safe to
run before a oneshot/update-pvc-daemons.yml playbook run.
2021-11-15 10:51:38 -05:00
Joshua Boniface 414678f683 Fix a few more extraneous splits
Just use this_node if applicable, or the raw node.hostname.
2021-11-11 17:35:42 -05:00
Joshua Boniface 243c910d6d Unify and standardize inventory_hostname
This was causing some confusing conflicts, so create a new fact called
"this_node" which is inventory_hostname.split('.')[0], i.e. the short
name, and use that everywhere instead of an FQDN or true inventory
hostname.
2021-11-11 17:19:03 -05:00
Joshua Boniface 5797535997 Adjust documentation and behaviour of cpuset
1. Detail the caveats and specific situations and ref the documentation
which will provide more details.

2. Always install the configs, but use /etc/default/ceph-osd-cpuset to
control if the script does anything or not (so, the "osd" cset set is
always active just not set in a special way.
2021-09-29 20:49:00 -04:00
Joshua Boniface 81cf341c32 Install cset configs even if disabled
The setup script handles this instead.
2021-09-29 10:23:01 -04:00
Joshua Boniface 645249b57e Allow dynamic enabling/disabling of cset
Add a separate config to handle enable/disable on the system itself.
2021-09-29 10:21:47 -04:00
Joshua Boniface 8ac2a5ea0c Adjust default ceph.conf parameters
1. Remove an explicit OSD journal size, especially such a small one (no
clue why I ever added that...)

2. Add max scrubs, disable scrub during recovery, and set scrub sleep.

3. Add max backfills, tune recovery sleep to 0 to prioritize recovery.
2021-09-28 02:09:50 -04:00
Joshua Boniface 732bfe732c Add Ceph OSD cpuset tuning options
Allows an administrator to set CPU pinning with the cpuset tool for Ceph
OSDs, in situations where CPU contention with VMs or other system tasks
may be negatively affecting OSD performance. This is optional, advanced
tuning and is disabled by default.
2021-09-27 00:27:57 -04:00
Joshua Boniface 2e9d02ab52 Add additional CMK checks 2021-08-21 15:41:44 -04:00
Joshua Boniface b62731199f Add default features flag to ceph.conf generator
Coupled with the removal of explicit --image-features flags to the RBD
command in PVC itself, this ensures that only the two features supported
on kernel 4.19 are enabled by default.
2021-07-30 11:39:24 -04:00
Joshua Boniface 3adacf3107 Fix excessive whitespace 2021-07-18 22:13:09 -04:00
Joshua Boniface 10a1754285 Adjust package lists per Debian version 2021-07-18 18:36:58 -04:00
Joshua Boniface c0ad9740f4 Fix bootstrap collection path for Ceph 2021-06-29 17:52:21 -04:00
Joshua Boniface 120871ee45 Support both versions of psycopg2 and kazoo 2021-06-29 17:29:01 -04:00
Joshua Boniface 6396eaa5ff Ensure libvirtd restarts when unit changes 2021-06-22 03:15:23 -04:00
Joshua Boniface ec879f4e3c Add override custom libvirtd.service unit
This has no functional change on Buster, but on Bullseye this overrides
the stupid socket-based activation shenanigans that the default unit
tries to do, as well as the breaking replacement of the
/etc/default/libvirt variable names.
2021-06-22 03:15:23 -04:00
Joshua Boniface 4ccc23bd85 Add python3 version of psycopg2 explicitly 2021-06-22 03:15:23 -04:00
Joshua Boniface 238449904f Move some other tasks to bootstrap role
Avoids an issue where the pvcnoded service is stopped on non-bootstrap
runs.
2021-05-13 10:17:38 -04:00
Joshua Boniface 404751f695 Update relative path to bootstrap files 2021-04-08 14:04:56 -04:00
Joshua Boniface 82fa85834a Add libguestfs-tools to libvirt role deps 2021-03-15 13:39:37 -04:00
Joshua Boniface 1c05c8729f Fix incorrect systemd enabling in Patroni 2021-01-28 16:28:02 -05:00
Joshua Boniface 04ca8f73d2 Correct bugs during bootstrap
1. Ensure Zookeeper restarts and checks out successfully before
proceeding with other steps.
2. Make sure PVC itself doesn't start prematurely.
2021-01-28 13:32:36 -05:00
Joshua Boniface b7f251ea16 Retry pgsql bootstrap startup 6 times
This will sometimes fail, so retry it several times
2021-01-27 15:45:36 -05:00
Joshua Boniface 7b08610efa Retry msgr2 enabling 6 times
This will sometimes fail, so retry it several times
2021-01-27 14:13:09 -05:00
Joshua Boniface 98e3e39570 Remove libjemalloc package 2020-12-21 00:20:45 -05:00
Joshua Boniface de04105a38 Add tuning for Ceph OSDs 2020-12-21 00:20:45 -05:00
Joshua Boniface cb96ef4e7a Use new init command location
Command was renamed in the PVC CLI to facilitate other "task" actions
like backup/restore.
2020-11-24 12:22:34 -05:00
Joshua Boniface 934f73af0f Support using existing SSL certs on system
Add the additional pvc_api_ssl_cert_path and pvc_api_ssl_key_path
group_vars options, which can be used to set the SSL details to existing
files on the filesystem if desired. If these are empty (or nonexistent),
the original pvc_api_ssl_cert and pvc_api_ssl_key raw format options
will be used as they were.

Allows the administrator to use outside methods (such as Let's Encrypt)
to obtain the certs locally on the system, avoiding changes to the
group_vars and redeployment to manage SSL keys.
2020-08-26 14:11:14 -04:00
Joshua Boniface 2f2123b70e Rename remaining "pvc_prov" items to pvc_api 2020-08-25 13:01:48 -04:00
Joshua Boniface 6b8232d38e Use cluster_group variable for paths
Instead of trying to automagic this group out of the Ansible hostvars,
just make it explicitly defined in the group_vars to avoid any
confusion.
2020-08-06 13:20:14 -04:00
Joshua Boniface a488f62ef8 Ignore errors in bringing up bootstrap interfaces 2020-07-27 13:08:24 -04:00
Joshua Boniface b0186b85c2 Use correct syntax for init command 2020-04-06 15:19:49 -04:00