Upgrade to Archipelago v0.4

Upgrade Steps

This guide assumes an upgrade from the previous (v0.3.x) version of Archipelago. This guide does not support upgrading from an intermediate snapshot version of Archipelago.

If you plan to integrate Archipelago with Pithos, all Archipelago nodes must be upgraded first.

The following steps must be applied to each node that gets upgraded.

  1. Prepare the node.
  2. Evacuate the node.
  3. Stop Archipelago
  4. Install the Archipelago v0.4 packages
  5. Adjust the new config file.
  6. Remove the old config file
  7. Start Archipelago

1. Prepare the node

Each node that will be updated must be idle with respect to Archipelago. To achieve the above the administrator must make sure that neither him nor the upper service layers perform any kind of Archipelago action on the node.

In order to do so, the administrator can manually set each node to be upgraded as drained using the following command on the ganeti master:

# gnt-node modify --drained=True <node>

or set the whole cluster to drained using the following snf-manage command from the Cyclades service node:

# snf-manage backend-modify --drained=True <backend_id>

If upgrading from Archipelago version less that 0.3.5, the administrator must make sure that each mapped volume on the node is opened. An oneliner like the following should do the trick:

# for i in $(vlmc showmapped|grep -v image|awk '{print $2}'); do vlmc open $i ; done ;

Make sure that no errors are raised. If so, the administrator must resolve them manually.

2. Evacuate the node

For each node to be upgraded, the administrator must evacuate it from Archipelago VMs, by either live-migrating them or failing them over to an already upgrade node. Of course, there is an exception on the first node to be upgraded.

3. Stop Archipelago

Archipelago must be fully stopped before upgrading. Perform the following command to achieve it:

# archipelago stop

4. Install the Archipelago v0.4 packages

Archipelago consists of the following packages:

  • libxseg0: libxseg used to communicate over shared memory segments
  • python-xseg: python bindings for libxseg
  • python-archipelago: archipelago python module that includes archipelago and vlmc functionality.
  • archipelago: user space tools and peers for the archipelago management and volume composition.
  • archipelago-ganeti: ganeti ext storage scripts, that enable ganeti to provision VMs over archipelago.

You can install them by installing archipelago archipelago-ganeti.

# apt-get install archipelago archipelago-ganeti

Optionally if you want RADOS support for archipelago, you should install archipelago-rados package.

# apt-get install archipelago-rados

and also xseg-tools package in case you need to dig into the shared memory segment

# apt-get install xseg-tools

On the nodes that will host VMs, blktap-archipelago-utils from GRNET and the distro-provided blktap-dkms package must also be installed.

# apt-get install blktap-archipelago-utils blktap-dkms

5. Adjust the new config file

The new config file is located on /etc/archipelago/archipelago.conf. There is also a sample configuration file for RADOS installations located at /etc/archipelago/archipelago.conf.rados_example. You can copy that to the Archipelago configuration file location in order to adjust installation-specific options only.

Adjust the new config file to the deployment needs (e.g. set filed directories or radosd pools).

Notable new config option that should be configured is:

  • BLKTAP_ENABLED: Whether or not the blktap module should be used. Must be set to true for nodes that will host VMs.

Archipelago v0.4 creates a new system user and group called archipelago. By default the configuration file is set up to run as those users. If your are using Archipelago with filed special care is needed:

  • You must change the corresponding USER and GROUP values of the configuration file to root, and follow the supplementary procedure on the end of this upgrade guide.
  • You must make sure that the archipelago user and group have the same permissions on the NFS share accross all nodes. This means for example that archipelago UID and GID are consistent across all Archipelago nodes for NFSv3 or there is a proper name mapping for NFSv4.

6. Remove the old config file

After migrating your setting from the old config to the new one, you can safely remove it from your system.

# rm /etc/default/archipelago

7. Start Archipelago

After successfully configuring the new/upgraded Archipelago installation, start it.

# archipelago start

After a successfull start, you can undrain the node:

# gnt-cluster modify --drained=False <node>

If you have drained the whole cluster and successfully upgraded all the nodes, you can undrain it using the snf-manage command:

# snf-manage backend-modify --drained=False <backend_id>

Finalizing upgrade

After upgrading all Archipelago nodes, you have to take certain steps to finalize the upgrade.

Adjust NFS shares permissions

As already mentioned, Archipelago v0.4 creates the new archipelago system user and group. In this section, we describe how to adjust the permissions of the directories and files on the NFS shares that Archipelago is using in order to run Archipelago as archipelago:archipelago. If you are not using Archipelago over NFS, skip this section.

We will refer to Archipelago data directory as the directory that holds the Archipelago data. On new installations this would probably be /srv/archip.

Warning

If you are integrating with a previous Synnefo installation, you must make sure that both Archipelago and Pithos have access to Archipelago data. You should skip this section, and perform the steps that are described in the Synnefo upgrade notes.

1. Change Archipelago data group permissions

Ensure that every file and folder under the Archipelago data directory has correct permissions.

# find /srv/archip/ -type d -exec chmod g+rwxs '{}' \;
# find /srv/archip/ -type f -exec chmod g+rw '{}' \;

2. Change the Archipelago data group owner

Make archipelago group the group owner of every file under the Archipelago data directory.

# chgrp archipelago /srv/archip/
# find /srv/archip/ -type d -exec chgrp archipelago '{}' \;
# find /srv/archip/ -type f -exec chgrp archipelago '{}' \;

From now on, every file or directory created under the Archipelago data directory will belong to the archipelago group because of the directory sticky bit that we set on the previous step. Plus the archipelago group will have full read/write access because of the SET_GUID bit.

3. Change Archipelago user and group

Now we can change the Archipelago configuration on all Archipelago nodes, to run as archipelago:archipelago user and group, since it no longer requires root priviledges.

For each Archipelago node:

  • Stop Archipelago

    # archipelago stop
    
  • Change the USER and GROUP configuration option to archipelago user. The configuration file is located under /etc/archipelago/archipelago.conf

  • Start Archipelago

    # archipelago start
    

Pithos integration when using Filed

If you are using Pithos backed by Archipelago with filed, after having upgraded all Archipelago nodes and successfully installed the upgraded Pithos version, the following steps must also be followed.

1. Stop all Archipelago instances

On every node that runs Archipelago, perform the following:

# archipelago stop

Use the -f option if there are mapped volumes. Have in mind that during the time Archipelago is stopped, the VMs will appear frozen whenever they attempt to perform any disk I/O.

2. Enable Pithos object migration

Enable the pithos_migrate setting for all blockerm and blockerb peers on all nodes. Add the following line pithos_migrate=True on the blockerm and blockerb section of the configuration files.

3. Start all Archipelago instances

On every node that runs Archipelago, perform the following:

# archipelago start

Change Filed lock files location

If your installation does not rely on filed skip this section.

In previous Archipelago versions, lock files were placed along with the data files of blockerm. In Archipelago version 0.4 we set a distinct lock file directory for easier lock lookup.

0. Prerequisites

Make sure you have a common directory shared with all Archipelago nodes (e.g. /srv/archip/locks). The directory must be owned by the user and group Archipelago run as (default archipelago:archipelago) and both the user and the group must have read and write permissions.

1. Stop all Archipelago instances

On every node that runs Archipelago, perform the following:

# archipelago stop

Use the -f option if there are mapped volumes. Have in mind that during the time Archipelago is stopped, the VMs will appear frozen whenever they attempt to perform any disk I/O.

2. Set lock directory

Set the lock directory for all blockerm peers on all nodes. Add the following line lock_dir=/srv/archip/lock where /srv/archip/locks is the shared directory created on step 0.

3. Start all Archipelago instances

On every node that runs Archipelago, perform the following:

# archipelago start

Convert all volume mapfiles

Archipelago lazily upgrades the mapfiles to the latest version, when they are accessed. To make sure that all mapfiles have been upgraded to the latest version, the provided migration tool must be executed. The tool is located in /usr/share/archipelago/tools/finalize_upgrade_0.4. You can run it from any node with access to Archipelago. Make sure that it completes successfully.

It is advised, in order to avoid false alarms (e.g. a mapfile that failed to upgrade), to be idle wrt to Archipelago control operations.