resalloc v5.11

• Bugfixes

  • Issue #173 has been resolved. The resalloc server no longer allows the
    processes spawned by CleanUnknownWorker to run indefinitely. Instead, if
    these processes do not finish within 120 seconds, the server will
    terminate them. This 120s timeout should not be a significant problem as
    terminating processes only typically involves making a few "cloud API"
    calls.

resalloc v5.10

Packages use an RPM built-in mechanism for creating system users.

resalloc v5.9

• Enhancements

  • The Pool's page in the Web UI now includes a "totals" line that
    displays overall statistics summarized for all pools.

  • The Pool's Web UI page now also shows the DELETE_REQUEST numbers.

  • The Resource's Web UI page now renders the Data field as <code>,
    which properly displays newlines.

resalloc v5.8

• Bugfixes:

  • Show pool descriptions in a details tag. Otherwise the UX was especially
    bad on vertical monitors

resalloc v5.7

• New Features:

  • A description field for pools in the pools.yaml config is now
    recognized and displayed by the Web-UI (PR #167).

• Bugfixes:

  • The Manager now waits until PostgreSQL is running (PR #165).

resalloc v5.6

• Bugfixes:

  • The Watcher thread no longer fails in the case of random database
    exceptions, such as "database is locked." It now simply retries, ensuring
    the cluster remains well-maintained (cleanup of dead machines works
    properly).

  • The Web-UI now sorts the list of Pools alphabetically (PR #161).

resalloc v5.5

• Bugfixes:

  • Previous versions of the Resalloc Web-UI loaded the pools.yaml
    configuration file. In cases of configuration issues, they attempted to
    log problematic details into a log file owned by resalloc-server
    (var/log/resallocserver). Such attempts resulted in file permission
    problems. The Web-UI logic has been modified to no longer read
    pools.yaml and to rely solely on the database. In the future, we have
    to implement a completely separate configuration for the Web-UI and cease
    loading server.yaml, which should not be readable by Apache either, for
    instance.

resalloc v5.4

• Bugfixes:

  • The agent-spawner workers now ignore the output from shell command hooks
    (redirects them to /dev/null to avoid SIGPIPE)

  • The agent-spawner server now correctly base64-encodes the ticket data.

  • The agent-spawner package now requires up2date python3-copr-common for
    fixed get_redis_connection() method.

  • The resalloc's tags/tags_on_demand configuration was cleaned up to
    allow users to keep them unspecified, several useful warnings
    related to tag misconfiguration were added.

resalloc v5.3

- The helper scripts have been moved from resalloc-server to the
  resalloc-helpers (sub)package.  Cloud drivers that rely on those
  scripts might be usable without the resalloc-server package and its
  transitive dependencies.

resalloc v5.2

New in v5.2:

• New Features

  • Added a new convenience script, /bin/resalloc-wait-for-ssh. This script
    is useful for cloud drivers that want to actively wait until the VMs
    (resources) become available on SSH.

  • The /bin/resalloc-check-vm-ip now optionally accepts the $SSH_USER
    environment variable if checking root-SSH isn't possible. Users can use
    cmd_livecheck: SSH_USER=centos resalloc-check-vm-ip.

• Bug Fixes

  • Compatibility fix with SQLAlchemy v2.

  • The $AGENT_SPAWNER_RESOURCE_DATA variable content is now base64 encoded.

  • The resalloc-agent-spawner package now correctly specifies all the
    required runtime dependencies.

  • The survive_server_restart client option now ignores all possible
    server socket errors, avoiding unexpected client interrupts.