summaryrefslogtreecommitdiffstats
path: root/src/pybind/mgr/cephadm/schedule.py (follow)
Commit message (Collapse)AuthorAgeFilesLines
* mgr/cephadm: fix placement with label and host patternAdam King2024-03-011-0/+2
| | | | | | | | | Previously, when both the label and host pattern were provided, only the label was actually used for the placement Fixes: https://tracker.ceph.com/issues/64428 Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm: filter hosts that can't support VIP for ingressAdam King2023-08-041-2/+2
| | | | | | | | | Keepalive daemons need the host to have an interface on which they can set up their VIP. If a host does not have any interface that can work, we should filter it out Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm: lower log level when logging about related servicesAdam King2023-05-081-6/+1
| | | | | | | | | | | | This was recently added in https://github.com/ceph/ceph/commit/088d2c4205c599a7d4f2ce4de8e2af8e129adac8 and seems to work fine, but logging these things at info level spams the log as every single service every serve loop iteration is reporting on whether it has related daemons. This changes it to only log when it finds related daemons or when it prefers a host due to those related daemons, and do both of those at only debug level. Signed-off-by: Adam King <adking@redhat.com>
* Merge pull request #51226 from jsoref/spelling-orchestratorAdam King2023-05-031-1/+1
|\ | | | | | | | | orchestrator: Fix spelling Reviewed-by: Adam King<adking@redhat.com>
| * orchestrator: fix spelling errorsJosh Soref2023-04-261-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * a new * accommodated * adopted * appended * because * bootstrap * bootstrapping * brackets * classes * cluster * compatible * completely * confusion * daemon * daemons * dashboard * enclosure * existing * explicit * following * format * host * implementation * inferred * keepalived * kubic * maintenance * necessarily * necessary * network * notifier * octopus * permanent * presenting * related * see * snapshot * stateful * the * track * version * wasn't * weird Signed-off-by: Josh Soref <2119212+jsoref@users.noreply.github.com>
* | mgr/cephadm: prefer same hosts as related service daemons when picking ↵Adam King2023-05-011-0/+28
|/ | | | | | | | | | | | | | | | | | | | | | | | arbitrary hosts For now, just for linking ingress services and their backend services. The idea is if one, or both, of the ingress service and backend service is using a count, to try and get them to deploy their daemons on the same host(s). If the placements have explicit placements (not using count) we still stick to those placements regardless. This should enable something like specifying a host for the backend service and leaving the ingress placement as just "count: 1" and having the ingress service get on the same host as the backend service daemon. This is particularly useful for the keepalive-only (VIP but no haproxy) over NFS setup where the keepalive must share a host with the NFS to function, but will also be useful for other VIP only setups we may do in the future. Signed-off-by: Adam King <adking@redhat.com>
* pybind/mgr: use annotation style without commentKefu Chai2022-11-241-2/+2
| | | | | | | | | | | | | | | | | | | | | | * use annotation style without comment * add noqa to import statement to prevent flake8 from complaining those which import typing modules to silence the warnings like: ``` flake8 run-test: commands[0] | flake8 --config=tox.ini alerts balancer cephadm cli_api crash devicehealth diskprediction_local hello iostat localpool nfs orchestrator prometheus selftest cephadm/schedule.py:5:1: F401 'typing.Callable' imported but unused cephadm/schedule.py:8:1: F401 'ceph.deployment.service_spec.ServiceSpec' imported but unused cephadm/tests/fixtures.py:17:1: F401 'orchestrator.OrchResult' imported but unused orchestrator/module.py:4:1: F401 'typing.Set' imported but unused orchestrator/module.py:17:1: F401 'ceph.deployment.inventory.Device' imported but unused prometheus/module.py:17:1: F401 'typing.DefaultDict' imported but unused 6 F401 'typing.Callable' imported but unused ERROR: InvocationError for command /home/jenkins-build/build/workspace/ceph-pull-requests/src/pybind/mgr/.tox/flake8/bin/flake8 --config=tox.ini alerts ``` Signed-off-by: Kefu Chai <tchaikov@gmail.com>
* mgr/cephadm: allow binding to loopback for rgw daemonsRedouane Kachach2022-08-311-0/+8
| | | | | | Fixes: https://tracker.ceph.com/issues/57304 Signed-off-by: Redouane Kachach <rkachach@redhat.com>
* mgr/cephadm: fix handling of draining hosts with explicit placement specsAdam King2022-08-111-2/+5
| | | | | | | | | | | | | | | Basically, if you have a placement that explicitly defines the hosts to place on, and then add _no_schedule label to one of the hosts (which should cause all daemons to be removed from the host) cpehadm will simply fail to apply the spec, saying the host with the _no_schedule label is "Unknown". This is due to the fact that we remove hosts with the _no_schedule label from the pool of hosts the scheduler has to work with entirely. If we also provide the scheduler with a list of currently draining hosts, it can handle this better and the daemon can be drained off the host as expected. Fixes: https://tracker.ceph.com/issues/56972 Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm: fixing scheduler consistent hashingRedouane Kachach2022-06-291-12/+7
| | | | Signed-off-by: Redouane Kachach <rkachach@redhat.com>
* mgr/cephadm: offline host watcherAdam King2022-03-261-1/+2
| | | | | | | | | To be able to detect if certain offline hosts go offline quicker. Could be useful for the NFS HA feature as this requires moving nfs daemons from offline hosts within 90 seconds. Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm: Reschedule nfs daemons from offline hostsAdam King2022-03-221-0/+16
| | | | | | | | | In order to improve nfs availability, if there are other hosts we can place an nfs daemon on or if there is a host with a lower rank nfs daemon when a higher rank one is on an offline host, we should reschedule the nfs daemons Signed-off-by: Adam King <adking@redhat.com>
* cephadm: fix upgrade name already in useDaniel Pivonka2021-11-051-2/+2
| | | | Signed-off-by: Daniel Pivonka <dpivonka@redhat.com>
* mgr/cephadm: Fixed spawning ip addresses list for public network interface.AndrewSharapov2021-10-291-2/+2
| | | | | | | | | Eevery call of find_ip_on_host() actually duplicates the list of public ip addresses in self.networks, while it should NOT change it. As the result value of key mgr/cephadm/host.<hostname> in kv store becomes very large and may cause crash of ceph mgr. Signed-off-by: Andrew Sharapov <andrewshar@gmail.com>
* mgr/cephadm: make scheduler able to accomodate offline/maintenance hostsAdam King2021-08-051-3/+19
| | | | | | Fixes: https://tracker.ceph.com/issues/51027 Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm/schedule: assign/map ranksSage Weil2021-05-191-21/+117
| | | | | | | | | | If we are passed a rank_map, use it maintain one daemon per rank, where the ranks are consecutive non-negative integers starting from 0. A bit of refactoring in place() so that we only do the rank allocations on slots we are going to use (no more than count). Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: add rank[_generation] propertiesSage Weil2021-05-191-0/+6
| | | | | | | | | | DaemonDescription CephadmDaemonDeploySpec DaemonPlacement unit.meta get_unique_name() (we include it in the daemon_id) Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: simplifySage Weil2021-05-191-2/+1
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: make placement shuffle deterministicSage Weil2021-05-191-3/+7
| | | | | | | | | | | | hash(str) is non-deterministic, probably because it is using the internal object ID or something and not the string content? In any case, explicitly hash the string content and use that instead. Also, sort the input pre-shuffle to ensure that variations in the original host list ordering don't screw with the result. Signed-off-by: Sage Weil <sage@newdream.net>
* Merge pull request #41257 from liewegas/cephadm-filter-net-messageKefu Chai2021-05-131-3/+0
|\ | | | | | | | | mgr/cephadm/schedule: fix message Reviewed-by: Sebastian Wagner <sewagner@redhat.com>
| * mgr/cephadm/schedule: fix filter messageSage Weil2021-05-101-3/+0
| | | | | | | | | | | | | | | | This is now only used when scheduling mons. (Units now enable the kernel features needed instead of checking for them during placement.) Move the message to the filter itself. Signed-off-by: Sage Weil <sage@newdream.net>
* | mgr/cephadm: misc. autopep8 changesAdam King2021-05-051-1/+2
|/ | | | | | | | | Some of the file in mgr/cephadm are getting changed every time I want to run the mgr tox tests to check changes and it's inconvenient to have to check the files out every time Signed-off-by: Adam King <adking@redhat.com>
* mgr/cephadm: use per_host_daemon feature in schedulerSage Weil2021-04-131-1/+1
| | | | | | This only affects ingress, at least for now. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: add per_host_daemon_type supportSage Weil2021-04-131-7/+37
| | | | | | | | | | This will be used to schedule a per-host keepalived alongside other services. Implement this as a final stage for place() that puts one per host and also takes existing/stray daemons into consideration. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: include daemon_type in DaemonPlacementSage Weil2021-04-131-6/+18
| | | | | | Initially, this will always match the service_type. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: update list-networks to report interface names tooSage Weil2021-04-131-2/+4
| | | | | | | Also, minor fix in the ipv6 addr reporting: ignore networks that aren't in CIDR form (no /). Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: handle multiple ports per daemonSage Weil2021-04-131-18/+22
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: make DaemonPlacement print nicerSage Weil2021-03-171-3/+1
| | | | | | 'host(ip:port)' or 'host(*:port)' so we can show it to a user. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: choose an IP from a subnet listSage Weil2021-03-171-1/+24
| | | | | | | | | | | | | | | | | | Choose an IP from the subnet list provided by the ServiceSpec. A few caveats: - we ignore hosts that don't have IPs in the given subnet - the subnet matching is STRICT. That is, the CIDR name has to exactly match what is configured on the host. That means you can't just say 10/8 to match any 10.whatever addres--you need the exactly network on the host (e.g, 10.1.2.0/24). - If you modify a servicespec and change the networks when there are already deployed daemons, we will try to deploy the new instances on the same ports but bound to a specific IP instead of *. Which will fail. You need to remove the service first, or remove the old daemons manually so that creating new ones will succeed. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: match placement ip only combination with portSage Weil2021-03-161-4/+5
| | | | | | | | 1- We only have an IP to bind to if we also have a port, and 2- If we do, we want an exact match: if the DaemonPlacement has ip of None, then the DaemonDescription should have None too. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: dynamically assign ports for rgwSage Weil2021-03-151-6/+23
| | | | | | | Dynamically number ports for RGW instances, with the start port being the one configured on the service (or the default of 80 or 443). Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: only 1 port in DaemonPlacementSage Weil2021-03-151-4/+4
| | | | | | | | It would be weird to dynamically number multiple ports (although doable). But until we have plans to support something like that, no need to handle it here. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: return DaemonPlacement instead of HostPlacementSpecSage Weil2021-03-151-29/+66
| | | | | | | | Create a new type for the result of scheduling/place(). Include new fields like ip:port. Introduce a matches_daemon() to see whether an existing daemon matches a scheduling slot. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: remove unused methodsSage Weil2021-03-151-18/+0
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: handle bare 'count-per-host:NNN', fix commentsSage Weil2021-03-101-3/+4
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: remove Scheduler abstraction (for now at least)Sage Weil2021-03-101-39/+1
| | | | | | It no longer does anything except slice the array. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: calculate additions/removals in place()Sage Weil2021-03-101-58/+30
| | | | | | | | We already have to examine existing daemons to choose the placements. There is no reason to make the caller call another method to (re)calculate what the net additions and removals are. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: allow colocation of certain daemon typesSage Weil2021-03-101-63/+48
| | | | | | | For certain daemon types, we can deploy more than one per host (mds, rbd-mirror, rgw). Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: shuffle candidates, not final placementsSage Weil2021-03-101-4/+4
| | | | | | | Otherwise we may end out randomly doubling up on some hosts and none on others (when we have more than one placement per host). Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: pass per-type allow_colo to the schedulerSage Weil2021-03-101-0/+11
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: pass daemons, not get_daemons_funcSage Weil2021-03-101-3/+3
| | | | | | | | - simpler - many/most callers already have the daemon list, so we save ourselves duplicated effort Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: move host filtering into get_candidates()Sage Weil2021-03-101-34/+22
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm/schedule: respect count-per-hostSage Weil2021-03-101-8/+15
| | | | | | | | In the no-count cases, our job is simple: we have a set of hosts specified via some other means (label, filter, explicit list) and simply need to do N instances for each of those hosts. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: adjust deployment logic to allow multiple daemons per hostSage Weil2021-03-101-5/+13
| | | | | | If we are told to deploy multiple instances on a host, do it. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: do not worry about even # of monitorsSage Weil2021-03-101-24/+0
| | | | | | | | | | | | | | Ceph works just fine with an even number of monitors. Upside: more copies of critical cluster data Downside: we can tolerate the same number of down mons as N-1, and now we are slightly more likely to have a failing mon because we have 1 more that might fail. On balance it's not clear that have one fewer mon is any better, so avoid the confusion and complexity of second-guessing ourselves. Signed-off-by: Sage Weil <sage@newdream.net>
* mgr/cephadm: Module imported but unused (F401)Michael Fritch2021-02-051-1/+1
| | | | Signed-off-by: Michael Fritch <mfritch@suse.com>
* mgr/cephadm: Do not use variables named 'I', 'O', or 'l' (E741)Michael Fritch2021-02-051-9/+9
| | | | Signed-off-by: Michael Fritch <mfritch@suse.com>
* mgr/cephadm: Missing whitespace around arithmetic operator (E226)Michael Fritch2021-02-051-2/+3
| | | | Signed-off-by: Michael Fritch <mfritch@suse.com>
* mgr/cephadm: fix capitalization, level; drop elipses of log msgsSage Weil2021-02-021-5/+5
| | | | Signed-off-by: Sage Weil <sage@newdream.net>
* cephadm: HA for RGW endpointsDaniel-Pivonka2021-01-081-5/+22
| | | | | | | | | | Cephadm deploying keepalived and HAproxy for providing High availability for RGW endpoints Fixes: https://tracker.ceph.com/issues/45116 Signed-off-by: Daniel-Pivonka <dpivonka@redhat.com> Signed-off-by: Adam King <adking@redhat.com> Signed-off-by: Juan Miguel Olmo Martínez <jolmomar@redhat.com>
generated by cgit v1.2.3 (git 2.39.1) at 2025-01-19 04:22:21 +0100