lp:~troyanov/maas

Author: Anton Troyanov
Author Date: 2025-06-23 12:51:13 UTC

fix(region): ignore disconnected rackd on BMC VLAN

Original implementation checked if there are any Rack Controllers that are
connected to BMC's VLAN and if such Rack Controllers exist, VLAN specific task
queue was used.

This fix introduces an extra filtering to ensure that disconnected Rack
Controllers are removed and they do not participate in power management.

Resolves LP:2115176

Author: Anton Troyanov
Author Date: 2025-05-08 20:08:39 UTC

fix: use cloud-init daily

Use ppa:cloud-init-dev/daily

In order to validate the following fix:
https://github.com/canonical/cloud-init/pull/6167

Author: Anton Troyanov
Author Date: 2025-04-30 15:05:41 UTC

fixup! fix(power): Proxmox Verify SSL with system CA fails

Author: Anton Troyanov
Author Date: 2025-04-30 09:32:16 UTC

test(agent): remove flaky test

Author: Anton Troyanov
Author Date: 2025-04-29 19:46:08 UTC

fixup! perf(agent): new pool implementation for resolver

Author: Anton Troyanov
Author Date: 2025-04-28 12:17:03 UTC

fix: linter

Author: Anton Troyanov
Author Date: 2025-04-25 14:57:33 UTC

fix(agent): set default cache_size to 20GB

Resolves LP:2109360

Author: Anton Troyanov
Author Date: 2025-04-18 10:38:18 UTC

refactor(agent): currentChain -> chain

Each time I see "currentChain" I am thinking that there is also another
chain that might exist.

Author: Anton Troyanov
Author Date: 2025-04-18 08:29:55 UTC

refactor(agent): simplify resolver cache constructor

Author: Anton Troyanov
Author Date: 2025-04-18 16:13:39 UTC

fixup! fix(agent): single AUTHORITY and ADDITIONAL for same questions

Author: Anton Troyanov
Author Date: 2025-04-15 05:32:34 UTC

fix(deps): proper location of Build-Depends

> Note that the binary package relationship fields such as Depends appear in one
  of the binary package stanzas of the template control file, whereas the
  build-time relationships such as Build-Depends appear in the source package
  stanza of the template control file (which is the first section).

https://www.debian.org/doc/debian-policy/ch-relationships.html#syntax-of-relationship-fields

Author: Anton Troyanov
Author Date: 2025-04-14 17:54:50 UTC

fix(agent): 60s timeout when testing configure-cluster-service

Microcluster initialization might take a bit longer than 3s on a
slow machine. In such case this test fails with an error:

--- FAIL: TestConfigurationWorkflow (3.00s)
panic: test timeout: 3s

To prevent this, this change bumps the test timeout to 60 seconds.

Author: Anton Troyanov
Author Date: 2025-04-10 21:04:46 UTC

fixup! perf(agent): disable logging for resolver benchmarks

Author: Anton Troyanov
Author Date: 2025-04-10 18:16:04 UTC

refactor(agent): improve resolver/cache_test.go

Author: Anton Troyanov
Author Date: 2025-04-07 16:30:45 UTC

fix(power): Proxmox Verify SSL with system CA fails

Resolves LP:2104907

Author: Anton Troyanov
Author Date: 2025-04-07 08:23:45 UTC

fix(agent): deb packaging

Fix a deb packaging regression introduced in:
2681b11f2a939d9b24221a1a07b4f15a09e30037

Author: Anton Troyanov
Author Date: 2025-04-03 12:15:53 UTC

fix: MAAS KVM deploy with 24.04 fails sshd.service not found

sshd.service was an alias to ssh.service and eventually got removed in
24.04

Resolves LP:2104278

Author: Anton Troyanov
Author Date: 2025-04-03 11:09:55 UTC

fix(agent): snap incorrectly calculates cache_size

When running inside a snap, the calculated `cache_size` should use a
correct mount point.

Do this:
```
ubuntu@maas:~/maas$ df -H /snap/maas/x1
Filesystem Size Used Avail Use% Mounted on

/dev/nvme0n1p2 251G 160G 79G 67% /snap/maas/x1
```

Not this:
```
ubuntu@maas:~/maas$ df -H /
Filesystem Size Used Avail Use% Mounted on
default/containers/maas 28G 3.4G 24G 13% /
```

Resolves LP:2104838

Author: Anton Troyanov
Author Date: 2025-04-03 07:05:10 UTC

refactor(agent): remove redundant cond check

Author: Anton Troyanov
Author Date: 2025-04-02 21:19:39 UTC

chore(agent): switch back to a microcluster v2

Issue was closed in the upstream, no need to use the fork anymore.
https://github.com/canonical/microcluster/issues/350

Author: Anton Troyanov
Author Date: 2025-04-01 15:46:24 UTC

chore(ci): migrate to golangci-lint v2.0.2

With other projects migrating to v2 it would be easier to maintain a
shared configuration.

Author: Anton Troyanov
Author Date: 2025-03-31 20:41:58 UTC

Revert "refactor(agent): simplify(?) TestSession_Add"

This reverts commit 544bc84f88be4c354f495701a9c02fb40170d4c1.

Author: Anton Troyanov
Author Date: 2025-03-27 06:29:14 UTC

feat(agent): statically linking ftw

Author: Anton Troyanov
Author Date: 2025-02-11 20:55:09 UTC

chore: replace interface{} with any

Go 1.18 introduced the `any` type as an alias to `interface{}` which is
now considered to be a more commonly used across other projects. Lets
try to stay with current best practices.

Author: Anton Troyanov
Author Date: 2025-02-08 13:16:00 UTC

refactor

Author: Anton Troyanov
Author Date: 2025-01-29 08:26:39 UTC

wip

Author: Anton Troyanov
Author Date: 2025-01-18 13:51:33 UTC

chore(agent): use Go 1.22 (supported)

MAAS 3.6 requires Ubuntu 24.04 LTS (Noble Numbat) which has Go 1.22

https://packages.ubuntu.com/noble/devel/golang-go
(cherry picked from commit 5c655a42afb6febf29d7b54da764743d47775f6b)

Author: Anton Troyanov
Author Date: 2025-01-17 09:38:39 UTC

chore(agent): use Go 1.22 (supported)

MAAS 3.6 requires Ubuntu 24.04 LTS (Noble Numbat) which has Go 1.22

https://packages.ubuntu.com/noble/devel/golang-go

Author: Anton Troyanov
Author Date: 2025-01-17 09:05:37 UTC

fix: Workflow execution already started

When MAAS Agent starts it triggers configure-agent workflow, which
then triggers some child-workflows. Desired behaviour is to have only a
single instance of such a workflow running (and all the child ones), hence we
set ID Reuse Policy to WorkflowIDReusePolicy.TERMINATE_IF_RUNNING

However there is another invocation at metadataserver/builtin_scripts/hooks.py
which doesn't enforce ID Reuse Policy and doesn't set custom workflow
ID. That allows two configure-agent workflow to be executed at the same time,
and that leads to an error with a child workflow:
START_CHILD_WORKFLOW_EXECUTION_FAILED_CAUSE_WORKFLOW_ALREADY_EXISTS

Resolves LP:2095085

Author: Anton Troyanov
Author Date: 2025-01-15 14:27:46 UTC

fix: Workflow panic. Potential deadlock detected

> Potential deadlock detected

This error happens because workflow thread did not yield control via a Temporal
API call for over a second.

This change moves all the logic (that we thought is safe to execute in the
workflow directly) to be executed as a Local Activity.

Resolves LP:2085160

Author: Anton Troyanov
Author Date: 2024-10-08 14:03:53 UTC

fix(snap): use ppa:maas/3.5-next

Author: Anton Troyanov
Author Date: 2024-10-07 14:21:31 UTC

fix(agent): custom un/marshal for the Host type

Instead of omapi.Host use dhcp.Host for un/marshal

Author: Anton Troyanov
Author Date: 2024-10-04 08:27:33 UTC

fix(agent): custom un/marshal for the Host type

Host is being used on both Python and Go side.
In Python it is defined as:

@dataclass
class Host:
    ip: Optional[str]
    mac: Optional[str]
    hostname: Optional[str]

Which leads to certain issues with MAC address as there is no custom
UnmarshalText (like for net.IP). And since net.HardwareAddr is actually a
[]byte it will fail to Unmarshal from "ca:fe:ba:be:00:00" as it is expecting
base64-encoded string.

From https://pkg.go.dev/encoding/json:
> Array and slice values encode as JSON arrays, except that []byte encodes
> as a base64-encoded string, and a nil slice encodes as the null JSON value.

Author: Anton Troyanov
Author Date: 2024-10-03 14:26:29 UTC

refactor: remove ValidateDHCP RPC call

Author: Anton Troyanov
Author Date: 2024-09-01 20:13:07 UTC

feat(agent): tracing using opentelemetry

Author: Anton Troyanov
Author Date: 2024-09-01 18:50:57 UTC

feat(agent): add configurable pprof endpoint

Add new pprof endpoint to the MAAS Agent that can be configured to be enabled
or disabled based on a setting. This allows better control over performance
profiling in different environments.

Author: Anton Troyanov
Author Date: 2024-09-18 22:09:11 UTC

fixup! feat: add dhcp-config-via-file activity

Add atomicfile.WriteFile

Author: Jacopo Rota
Author Date: 2023-06-23 07:41:09 UTC

Do not add neighbour during deployment if an ip address is in use

(cherry picked from commit 0144c504d72cff6e4a02fc89ba26e0774fe0b31c)

Author: Anton Troyanov
Author Date: 2024-09-12 13:15:41 UTC

fixup! feat: internal API handler to return DHCP configuration

Add missing tests

Author: Anton Troyanov
Author Date: 2024-09-11 15:30:23 UTC

fix: lint-go-fix Makefile target

Author: Anton Troyanov
Author Date: 2024-09-10 21:46:21 UTC

refactor(agent): extract service controller

Author: Anton Troyanov
Author Date: 2024-09-08 11:36:29 UTC

feat(maasapiserver): agents/{system_id/services/{service_name}/config

Author: Anton Troyanov
Author Date: 2024-09-04 21:40:38 UTC

fix(maasagent): use net.JoinHostPort

Concatenate a host and a port into a properly formatted address,
For IPv6 it will return [::1]:8080 instead of ::1:8080

Author: Anton Troyanov
Author Date: 2024-09-03 07:55:32 UTC

fixup! chore(maasapiserver): add ConfigurationsService.get_many

Author: Anton Troyanov
Author Date: 2024-09-02 18:23:00 UTC

chore: keep tables definitions ordered

Author: Anton Troyanov
Author Date: 2024-08-30 15:12:56 UTC

test(maasapiserver): fix TestConfigurationsService

The idea of the *TestSuite is to have abstract tests that would run on the
classes that extend it.

ConfigurationsService is a test on its own, not an abstract test.

Author: Anton Troyanov
Author Date: 2024-08-30 15:06:07 UTC

test: pytest should run *TestSuite classes

Author: Anton Troyanov
Author Date: 2024-08-28 15:00:08 UTC

fixup: specify IP for dummy0 with TEST_OMAPI_CLIENT_DUMMY_IP

Author: Anton Troyanov
Author Date: 2024-08-27 22:25:01 UTC

chore: remove redundant double quote

Author: Anton Troyanov
Author Date: 2024-08-27 17:06:54 UTC

test(agent): add testing/lxd package

testing/lxd is a Go package that provides test helper that allows
creation of LXD containers, which can be used for integration testing.

Author: Anton Troyanov
Author Date: 2024-08-12 14:40:58 UTC

Revert "Prepare for 3.3.8 release"

This reverts commit fdca268699a66cdc1a51fb4b93a20d0dba68027c.

Author: Anton Troyanov
Author Date: 2024-08-12 12:53:03 UTC

Prepare for 3.3.8 release

Author: Anton Troyanov
Author Date: 2024-08-11 19:23:39 UTC

refactor: unused context.Context

Author: Anton Troyanov
Author Date: 2024-08-11 09:58:32 UTC

refactor: use context.TODO() in tests

TODO returns a non-nil, empty Context.
Code should use context.TODO when it's unclear which Context to use or it is not
yet available (because the surrounding function has not yet been extended to
accept a Context parameter).

Author: Anton Troyanov
Author Date: 2024-08-09 14:29:54 UTC

refactor(agent): changes to dhcpd lease handler/storage

Add intermediate buffered channel, so reader and writer do not block
each other while there buffer is not full.

Rename table tests look consistent across MAAS Go codebase.

Author: Anton Troyanov
Author Date: 2024-08-07 09:44:17 UTC

feat(agent): DHCP service skeleton

This commit adds MAAS Agent DHCP service skeleton. Follow-up commits
will add glue to integrate it with ISC DHCP in the same way as it
currently works with the Rack Controller (listen over UNIX socket for
events and dispatch information to the Region Controller).

Author: Anton Troyanov
Author Date: 2024-08-01 17:38:49 UTC

fix(agent): use workflowID reuse policy for configure-agent

Provide custom workflow ID and use WORKFLOW_ID_REUSE_POLICY_TERMINATE_IF_RUNNING

Sometimes `configure-agent` workflow after series of quick restart of
maas-agent might throw an error "Workflow execution already started".

That happens because child workflow like `configure-httpproxy-service` or
`configure-power-service` have custom IDs and no workflow reuse policy,
making it impossible to start a new one, before the existing one times
out.

Author: Anton Troyanov
Author Date: 2024-08-01 17:26:55 UTC

refactor(agent): allow workflows and activities on main worker

Change Configurator interface, so it is possible to pass multiple
workflows and activities that can be registered on the main worker of
the worker pool.

Author: Anton Troyanov
Author Date: 2024-07-27 12:55:40 UTC

feat(agent): add file cache metrics

HTTP proxy is using file cache for boot resources.
Expose cache hits/misses/errors for better o11y.

Author: Anton Troyanov
Author Date: 2024-07-31 15:06:35 UTC

refactor(agent): improve errors for http proxy

Author: Anton Troyanov
Author Date: 2024-07-28 11:10:52 UTC

feat(agent): enable otel temporal client metrics

Author: Anton Troyanov
Author Date: 2024-07-28 11:08:51 UTC

chore(deps): bump go.temporal.io/sdk

Author: Anton Troyanov
Author Date: 2024-07-27 20:59:21 UTC

feat(agent): otel metrics with prometheus exporter

Expose MAAS Agent Go Runtime metrics using OpenTelemetry and Prometheus exporter.

Author: Anton Troyanov
Author Date: 2024-07-27 21:02:30 UTC

chore(deps): bump github.com/golang/protobuf

https://github.com/protocolbuffers/protobuf/issues/16163

Author: Anton Troyanov
Author Date: 2024-07-27 12:41:29 UTC

chore(ci): bump golangci-lint version v1.59.1

Author: Anton Troyanov
Author Date: 2024-07-27 12:03:57 UTC

chore(deps): bump github.com/stretchr/testify v1.9.0

Author: Anton Troyanov
Author Date: 2024-07-27 12:16:54 UTC

refactor: rand.Seed is deprecated since Go 1.20

SA1019:
  rand.Seed has been deprecated since Go 1.20 and an alternative has been
  available since Go 1.0: As of Go 1.20 there is no reason to call Seed with a
  random value. Programs that call Seed with a known value to get a specific
  sequence of results should use New(NewSource(seed)) to obtain a local random
  generator. (staticcheck)

Author: Anton Troyanov
Author Date: 2024-07-28 11:10:52 UTC

feat(agent): temporal otel metrics

Author: Anton Troyanov
Author Date: 2024-07-25 21:51:15 UTC

WIP

Author: Anton Troyanov
Author Date: 2024-07-17 10:08:57 UTC

feat(power): HMC Z option to set TLS verification

Resolves LP:2056208

Author: Anton Troyanov
Author Date: 2024-07-19 08:41:41 UTC

fix: maas-regiond should require maas-apiserver

Resolves LP:2073540

(cherry picked from commit aba348e01655f1eef3fe59613d90f6709009d109)

Author: Anton Troyanov
Author Date: 2024-07-18 20:29:36 UTC

fix: maas-regiond should require maas-apiserver

Resolves LP:2073540

Author: Anton Troyanov
Author Date: 2024-06-13 06:46:23 UTC

fixup! refactor: agent main, make it easier to follow

Author: Anton Troyanov
Author Date: 2024-06-07 07:12:26 UTC

fix(api): allow_dns/allow_proxy accept int and bool

Documentation mentions that allow_dns/allow_proxy parameters can be configured
using boolean or integer.

Resolves LP:2068666

(cherry picked from commit 04e4b65caa2906fb0b029e7510203d28a9d280d7)

Author: Anton Troyanov
Author Date: 2024-06-06 21:17:38 UTC

fix(api): allow_dns/allow_proxy accept int and bool

Documentation mentions that allow_dns/allow_proxy parameters can be configured
using boolean or integer.

Resolves LP:2068666

Author: Anton Troyanov
Author Date: 2024-06-05 08:15:17 UTC

fixup! refactor: refactor agent main

Author: Anton Troyanov
Author Date: 2024-06-03 06:54:34 UTC

fix: return 0 bytes for non-existing .lst

Resolves LP:2067793

(cherry picked from commit 54f3e5e544d998465ca7fd94b5c3e74b58ccc097)

Author: Anton Troyanov
Author Date: 2024-06-02 12:46:39 UTC

WIP

Author: Anton Troyanov
Author Date: 2024-06-01 10:57:03 UTC

fix: return 0 bytes for non-existing .lst

Resolves LP:2067793

Author: Anton Troyanov
Author Date: 2024-05-30 19:06:08 UTC

fix(cli): maas login can't handle non-ascii cacert

Resolves LP:2067503

(cherry picked from commit 2c935cce3833e308bfbeb15c5e00869ca1f4b8e2)

Author: Anton Troyanov
Author Date: 2024-05-29 20:35:47 UTC

fix(cli): maas login can't handle non-ascii cacert

Resolves LP:2067503

Author: Anton Troyanov
Author Date: 2024-05-29 22:28:09 UTC

fix(agent): create runDir if it doesn't exist

(cherry picked from commit f500d6ccfbbdcbdc09674cf97114991a9bcf01be)

Author: Anton Troyanov
Author Date: 2024-05-27 20:05:41 UTC

refactor: change Configure() to return interface{}

This allows to use Configurator interface when registering services,
however each service can define it's own configure method, so we can
pass required parameters when calling the workflow from the Region,
instead of having each service to make a callback via activities.

Author: Anton Troyanov
Author Date: 2024-05-29 19:55:37 UTC

fix(agent): create runDir if it doesn't exist

Author: Anton Troyanov
Author Date: 2024-05-29 19:21:56 UTC

Revert "fix(snap): create /run/snap.maas with install hook"

This reverts commit 2923235dfc975a11977460d023b32d2b49a14ff4.

Author: Anton Troyanov
Author Date: 2024-05-29 19:18:42 UTC

Revert "fix(snap): create /run/snap.maas with install hook"

This reverts commit d4c47cccef119d0a640be6e0134fc16feec5a282.

Author: Anton Troyanov
Author Date: 2024-05-29 17:51:32 UTC

fix(snap): create /run/snap.maas with install hook

Currently this folder is created by `chrony`, because `chrony` will create a
full path as defined for the PID file in the config [0]

We should not rely on `chrony` but rather create if after snap install,
because other services are using it (e.g. maas-agent)

```
if "SNAP" in os.environ:
    run_dir = get_maas_run_path()
    lines.append(f"dumpdir {run_dir / _NTP_DIR_NAME}")
    lines.append(f"pidfile {run_dir / _NTP_PID_NAME}")
    lines.append(f"bindcmdaddress {run_dir / _NTP_SOCK_NAME}")
```

[0]: src/provisioningserver/ntp/config.py

(cherry picked from commit d4c47cccef119d0a640be6e0134fc16feec5a282)

Author: Anton Troyanov
Author Date: 2024-05-29 15:12:23 UTC

fix(snap): create /run/snap.maas with install hook

Currently this folder is created by `chrony`, because `chrony` will create a
full path as defined for the PID file in the config [0]

We should not rely on `chrony` but rather create if after snap install,
because other services are using it (e.g. maas-agent)

```
if "SNAP" in os.environ:
    run_dir = get_maas_run_path()
    lines.append(f"dumpdir {run_dir / _NTP_DIR_NAME}")
    lines.append(f"pidfile {run_dir / _NTP_PID_NAME}")
    lines.append(f"bindcmdaddress {run_dir / _NTP_SOCK_NAME}")
```

[0]: src/provisioningserver/ntp/config.py

Author: Anton Troyanov
Author Date: 2024-05-29 09:01:47 UTC

fix: re-add missing foreign key constraints

f7230326051628a654afe8e2383be6faec05f35e had side-effect that removed
most of the foreign key relations

Resolves LP:2066936

Original fix was provided in 91490a921014dda26edec94689910148de8fc523,
but this commit excludes merge migration.

Author: Anton Troyanov
Author Date: 2024-05-28 21:12:11 UTC

fixup! Fix test_Commission_starts_commissioning intermittent failure.

Author: Anton Troyanov
Author Date: 2024-05-04 08:55:05 UTC

chore(deps): bump pebble to 1.10.2

Resolves LP:2064726

(cherry picked from commit 012430e452b8bd4e1714de51c7d11bb5c5b70e7e)

Author: Anton Troyanov
Author Date: 2024-05-04 06:54:44 UTC

 chore(deps): bump pebble to 1.10.2

Author: Anton Troyanov
Author Date: 2024-04-30 08:58:11 UTC

refactor: use configure-agent global timeout

Use WorkflowExecutionTimeout of configure-agent and let Temporal retry
all the child workflows and activities within this timeout.

Author: Anton Troyanov
Author Date: 2024-04-26 18:51:09 UTC

fix(tftp): legacy BIOS mode fails with 0-size cfg

The fix for GRUB introduced in a0b47aba48b2ad324c34b42aed9f47a7dfc85b28
lead to an issue with legacy BIOS mode, which is expecting to receive
`Error code 1: File not found` for non existing pxelinux.cfg config
files instead for a 0-size result.

Resolves LP:2063844

(cherry picked from commit 915882cfd7786a68d10a5cf461a29d7690912b80)

Author: Anton Troyanov
Author Date: 2024-04-26 15:42:25 UTC

fix(tftp): legacy BIOS mode fails with 0-size cfg

The fix for GRUB introduced in a0b47aba48b2ad324c34b42aed9f47a7dfc85b28
lead to an issue with legacy BIOS mode, which is expecting to receive
`Error code 1: File not found` for non existing pxelinux.cfg config
files instead for a 0-size result.

Resolves LP:2063844

Author: Anton Troyanov
Author Date: 2024-04-24 16:37:40 UTC

fix(agent): check if region endpoints are reachable

Region Controller might return IP:PORT that is not reachable by the
Agent (due to firewall configuration or network setup). Hence we should
add check if endpoints are reachable before applying them as targets
when configuring Agent HTTP reverse proxy service.

Resolves LP:2063220

Workflow logic is constrained by deterministic execution requirements.
In Go, Workflow Definition code cannot directly do the following:
- Call an external API, conduct a file I/O operation, talk to another service, etc

Normally we should use Activity for this scenario, however we are
confident and consider that we want to do this I/O inside the workflow.

But by having a blocking I/O you might get the following panic:
```
panic: Potential deadlock detected: workflow goroutine "root" didn't yield for over a second
```

It happens simply because the "root" goroutine was not scheduled for
execution. By using `workflow.Go` we don't mess with the scheduler and
we should worry about logic inside `workflow.Go` not take more than 1
second.

(cherry picked from commit 17596243156a800d060b8c5138798da8eae0950b)

Author: Anton Troyanov
Author Date: 2024-04-24 12:12:32 UTC

test(agent): region endpoints are unreachable

Author: Anton Troyanov
Author Date: 2024-04-05 22:12:14 UTC

fix: nginx rackd.conf httpproxy socket location

Resolves LP:2060288

(cherry picked from commit 2c575e55a3f007c37c38f0d2933921783804f642)

Author: Jacopo Rota
Author Date: 2024-04-05 13:09:11 UTC

fix: lp-2060278. Rename image sync lock file to match other maas lock files

(cherry picked from commit 668b2cd9f24661520c022bc11f74286cd1d1bc01)

Author: Alessandro Marcolini
Author Date: 2024-04-05 14:16:27 UTC

Fix lp bug 2033632
https://bugs.launchpad.net/maas/+bug/2033632

(cherry picked from commit 147ce845b0575adf944eccafff804caf0bdd876e)