1. 07 Aug, 2020 4 commits
  2. 04 Aug, 2020 2 commits
  3. 31 Jul, 2020 2 commits
    • Mike Hibler's avatar
      Speed up the recognition that a sync is done. · e50ad975
      Mike Hibler authored
      When an aggregate experiment is detected as "dead", we no longer restart
      syncthing to turn off the file watcher. By restarting syncthing, the new
      instance has to make a full scan of the directory even if it was INSYNC
      before. That can delay notification to the user by one or two monitor
      passes (10 minutes) or more.
      So if the syncthing is already running when the experiment transitions
      to dead, we leave it running with the watcher enabled. If no syncthing
      was running, we start it without the file watcher since a full scan will
      be needed anyway.
      Combined with shortening the rescan interval of the monitor to two
      minutes, we should now be able to get a tarball made and user notification
      done in under 5 minutes for the case where the save directory was already
      in sync when the experiment terminated.
    • Mike Hibler's avatar
  4. 29 Jul, 2020 2 commits
  5. 28 Jul, 2020 12 commits
  6. 27 Jul, 2020 3 commits
  7. 26 Jul, 2020 2 commits
  8. 22 Jul, 2020 6 commits
    • David Johnson's avatar
      Ensure that systemd-networkd dhcp client does not set a hostname. · 13b7f95f
      David Johnson authored
      (Since our dhcp servers only return node_id in the hostname field,
      we don't really want a dhcp client to be allowed to set the hostname,
      at least in the world of systemd-networkd.  The reason is that our
      systemd-networkd hooks/wrappers have to restart systemd-networkd once
      we've detected which device is the control net, since systemd-networkd
      cannot reload its configuration.  So we don't want to give it a second
      change to re-apply its dhcp-learned config (including hostname)
      because it might race with our sethostname.dhclient.
    • Leigh B Stoller's avatar
      Fix some problems with admission control wrt individually reservable · 574de3ea
      Leigh B Stoller authored
      * Since node_ids are now in the reservation (as the type), watch for
        this and do admission control on the node, previously we were letting
        this get to assign, where the error is, well, obscure.
      * Related to above, when asking for a reservable node, check to see if
        its already allocated before letting it get to assign. Better error
        message, and faster to return to the user.
      * Attempt to deal with a backwards compatibility issue wrt the
        matrix (but is a general issue); there are numerous profiles that ask
        for matrix nodes (or mmimo1, iris1, iris2) by their type since that is
        how they were reserved. But when we changed them to reservable nodes,
        existing profiles that ask for them by type suddenly broke cause
        1) the user wants a specific nuc, not any nuc, and 2) say in the case
        of mmimo1, there was only one node of that type but now admission
        control thinks there are no nodes of that type since individually
        reservable overrides that.
    • Leigh B Stoller's avatar
      Fix problem when using P2PLTE links in an experiment using virt · ba5726ab
      Leigh B Stoller authored
      nodes (or more generally, multiplexed links); force link encap to
      default, emulated to off, nobwshaping to off.
    • Leigh B Stoller's avatar
      Forgot this commit; change to type lists resulting from the addition of · ab38ef7f
      Leigh B Stoller authored
      individually reservable nodes we added a couple of weeks back,
    • Leigh B Stoller's avatar
      Minor fix. · 06719b4b
      Leigh B Stoller authored
    • Leigh B Stoller's avatar
      Work on issue #569; allow PI/Managers to delete/modifyinfo datasets in · f0280dc2
      Leigh B Stoller authored
      their project (subgroup).
  9. 21 Jul, 2020 2 commits
  10. 20 Jul, 2020 4 commits
  11. 17 Jul, 2020 1 commit