1. 01 Dec, 2014 1 commit
    • Dimitris Aragiorgis's avatar
      deploy: Networking fixes · da63a421
      Dimitris Aragiorgis authored
      Introduce /etc/network/interfaces.synnefo that will setup bridges
      used for VMs' public and private networks.
      
      Instead of using a common bridge use two separate bridges (br0,
      prv0). Their default ports are defined in nodes.conf
      (vm_public_iface, vm_private_iface) which are by default dummy. In a
      single node setup we don't need those interfaces. In case of router
      node, br0 will get the gateway IP.
      
      Use ferm for persistent network configuration. Add masq.ferm that
      is responsible for masquerading on router node and MAC SNAT on all
      nodes (needed in case we have a MAC filtered setup on
      vm_public_iface).
      
      Use rc.local to losetup extra disk and ifup bridges.
      Signed-off-by: default avatarDimitris Aragiorgis <dimara@grnet.gr>
      da63a421
  2. 30 Nov, 2014 2 commits
  3. 28 Nov, 2014 1 commit
  4. 26 Nov, 2014 7 commits
    • Olga Brani's avatar
      ui: Fix trailing comma in Javascript files · 184a99a0
      Olga Brani authored
      184a99a0
    • Olga Brani's avatar
      ui: Javascript issues with trailing comma in IE7 · a44a5290
      Olga Brani authored
      Fixes grnet/synnefo#320
      a44a5290
    • Olga Brani's avatar
      ui: Change default pagination in astakos projects list view · 423b3715
      Olga Brani authored
      Fixes grnet/synnefo#319
      423b3715
    • Stratos Psomadakis's avatar
      psycopg: Don't include conn info in exception · e0ae919f
      Stratos Psomadakis authored
      When trying to get a pooled_psycopg2 connection, the _get_pool() method
      checks the connection args against the pool args and raises an exception
      if the two differ. This leads to an internal server error, which results
      in logging the connection and pool info / args, which include the DB
      password, and, depending on the deployment, it might also send them via
      email to the admins. Log the connection and pool args only when log
      level is set to  debug, and remove them from the exception message, so
      that they won't end up being send via email.
      e0ae919f
    • Stratos Psomadakis's avatar
      docs: Mention admin-app pooling incompatibility · 3d37e58c
      Stratos Psomadakis authored
      snf-admin-app opens connections to multiple DBs, which means it cannot
      use psycopg connection pooling. Advise the users to delete any
      pooling-related options from 10-snf-webproject-database.conf (if, for
      example, they copied it from the Astakos or the Cyclades node).
      3d37e58c
    • Chrysostomos Nanakos's avatar
      webproject: Extend allocated xseg pool array · bd8e07f6
      Chrysostomos Nanakos authored
      On gunicorn reload, extend allocated xseg pool per worker class
      array until the already spawned workers exit. On a graceful worker
      exit the previously allocated xseg pool is properly shutdown and
      the new workers takeover the next series of xseg ports.
      On a next reload basis the new workers will always take over the
      previously allocated xseg pool ports. With this mechanism we always
      need at least twice the number of workers in xseg ports per pool.
      
      Care must be taken in xseg ports accounting and bookkeeping when
      we dynamically change the number of workers by using the TTIN and
      TTOUT signals sent to the master.
      
      The total number of xseg ports allocated by all workers must not
      exceed the minimum port being used by the Archipelago peers.
      If the total amount of used ports exceeds this limit one should
      rearrange the Archipelago peers ports along with the relevant
      dynamic port range.
      bd8e07f6
    • Stratos Psomadakis's avatar
      Merge pull request #318 from Erethon/hotfix-0.16.1 · caef4476
      Stratos Psomadakis authored
      docs: Update docs to comply with Ganeti 2.10
      caef4476
  5. 25 Nov, 2014 1 commit
  6. 24 Nov, 2014 7 commits
  7. 21 Nov, 2014 9 commits
  8. 19 Nov, 2014 1 commit
    • Christos Stavrakakis's avatar
      cyclades: Fix wait-for-sync on stopped instances · 30fd1f7f
      Christos Stavrakakis authored
      Pass 'wait_for_sync' option only on started Ganeti instances, because Ganeti
      will fail with 'OpPrereqError' in case the 'wait_for_sync' is set to
      False and the instance has deactivated disks.
      
      Note that Cyclades will not let a user neither attach a volume to an instance
      that is is in transition state, e.g. stopping, nor change the state of
      an instance if there is a pending volume attachment. However, in case
      the administrator manually modifies the instance via Ganeti, there is a
      race between checking the state of the instance and the time that Ganeti
      job will run that cannot be prevented.
      
      Closes grnet/synnefo#314
      30fd1f7f
  9. 18 Nov, 2014 2 commits
  10. 11 Nov, 2014 3 commits
  11. 10 Nov, 2014 3 commits
  12. 06 Nov, 2014 3 commits