Difference between revisions of "Testbed Update Plan"

From PDP/Grid Wiki
Jump to navigationJump to search
Line 14: Line 14:
  
 
Currently, using plain libvirt seems to fit most of our needs.
 
Currently, using plain libvirt seems to fit most of our needs.
 
The machines blade13 and blade14 are setup with Debian squeeze, libvirt and KVM with a Fiber Channel link to the compellent storage over which clustered LVM is defined to share access to the pool.
 
 
The machines arrone and aulnes are setup with Debian squeeze, libvirt and KVM with an NFS storage backend on put.testbed (storage.testbed) over which clustered LVM is defined to share access.
 
 
It happens that on reboot of a machine you need to manually start the pool and/or run vgscan. It is unclear at this moment why this would happen.
 
virsh pool-start vmachines
 
 
  error: Failed to start pool vmachines
 
  error: internal error '/sbin/vgchange -ay vmachines' exited with non-zero status 5 and signal 0:  Error locking on node arrone.testbed: Volume group for uuid not found: hI7udF9MvGpKkvkympcNG42glpwtwDeqhv7xV9wKHHdv9tDxQ9j8Lhgqem1esMcA
 
 
The following sequence of commands can be useful:
 
vgscan
 
 
  Reading all physical volumes.  This may take a while...
 
  Found volume group "vmachines" using metadata type lvm2
 
 
virsh pool-start vmachines
 
 
  Pool vmachines started
 
 
virsh pool-list --all
 
 
  Name                State      Autostart
 
  -----------------------------------------
 
  default              active    yes     
 
  put.testbed          active    yes     
 
  vmachines            active    yes
 
 
Also check the state of the LVM cluster, logs are in /var/log/cluster/. In the (rare) case the fence daemon has blocked the use of the cluster, the '''other''' machine in the cluster must be told it is ok to release the block. What the other machine is can be found in /etc/cluster/cluster.conf.
 
fence_ack_manual other-machine
 
  
 
=== Installing clustered lvm for live migration ===
 
=== Installing clustered lvm for live migration ===

Revision as of 16:26, 8 January 2015

Planning the update of the middleware/development test bed

The upgrade has taken place; the remaining information of this page is to be merged into Agile testbed.




Migration to a cloud infrastructure

Previous testbed cloud experiences are reported here.

Currently, using plain libvirt seems to fit most of our needs.

Installing clustered lvm for live migration

The machines arrone.testbed and aulnes.testbed, running Debian stable, are now equipped with a clustered LVM setup with an iSCSI device as the backend storage. The iSCSI device is served by put.testbed running FreeNAS.

Documentation to set up CLVM is available for Red Hat and Debian, both should be comparable.

First, a cluster needs to be defined and all systems in the cluster need to use the same definition. Put the file in /etc/cluster/cluster.conf:

<cluster name="vmachines" config_version="1">
  <cman expected_votes="1" two_node="1">
  </cman>
  <clusternodes>
    <clusternode name="arrone.testbed" votes="1" nodeid="1">
      <fence>
      </fence>
    </clusternode>
    <clusternode name="aulnes.testbed" votes="1" nodeid="2">
      <fence>
      </fence>
    </clusternode> 
  </clusternodes>
  <logging to_syslog="yes" to_logfile="yes" syslog_facility="daemon" syslog_priority="info">
  </logging>
  <fence_daemon post_join_delay="30" />
  <totem rrp_mode="none" secauth="off"/>
</cluster>

The setting 'two_node' is a special case for two node clusters, because there is no sensible way to do majority voting. In case one of the machines fails, the other will block to fence the first machine (which is a manual operation in our case) but the cluster can carry on with just one machine if needs be.

The machines keep an eye on one another through multicast, and therefore it is important to remove the following line from /etc/hosts:

127.0.1.1 arrone.testbed

which the Debian installation inserted. This makes the cluster manager daemon bind the wrong device for multicasts (the loopback device).

Another snag found on installation is the missing directory /var/run/lvm, which causes the startup script of clvm to fail. Once this is fixed, run

/etc/init.d/cman start
/etc/init.d/clvm start

Finally, the file

/etc/lvm/lvm.conf

needs to be edited to set

locking_type = 3

in order to use clustered locking.

Through this shared storage it is possible to do live migration of virtual machines between arrone and aulnes.

Installing Debian on blade 13 and 14 with Fiber Channel

This is a quick note to record a recent quirk. Although FC support on Debian works fine, using the multipath-tools-boot package is a bit tricky. It will update the initrd to include the multipath libraries and tools, to make it available at boot time.

This happened on blade-13; on reboot it was unable to mount the root partition (The message was 'device or resource busy') because the device mapper had somehow taken hold of the SCSI disk. By changing the root=UUID=xxxx stanza in the GRUB menu to root=/dev/dm-2 (this was guess-work) I managed to boot the system. There were probably several remedies to resolve the issue:

  1. rerun update-grub. This should replace the UUID= with a link to /dev/mapper/xxxx-part1
  2. blacklist the disk in the device mapper (and running mkinitramfs)
  3. remove the multipath-tools-boot package altogether.

I opted for blacklisting; this is what's in /etc/multipath.conf:

blacklist {
 wwid 3600508e000000000d6c6de44c0416105
}


A bonnie test of a VM with disk on local disk vs. a VM with disk on FC:

  • lofarwn.testbed had its disk locally
Version 1.03e       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
lofarwn          4G 20869  35 30172   5 23198   6 45957  85 510971  84 +++++ +++
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16 16759  99 +++++ +++ +++++ +++ 16810 100 +++++ +++ +++++ +++
lofarwn,4G,20869,35,30172,5,23198,6,45957,85,510971,84,+++++,+++,16,16759,99,+++++,+++,+++++,+++,16810,100,+++++,+++,+++++,+++
  • ige-cert.testbed has disk on LVM via FC.
Version 1.03e       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
ige-cert         4G 53384  96 216611  37 102283  24 51060  95 689474  79 +++++ +++
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16 12676 100 +++++ +++ +++++ +++ 12761  99 +++++ +++ +++++ +++
ige-cert,4G,53384,96,216611,37,102283,24,51060,95,689474,79,+++++,+++,16,12676,100,+++++,+++,+++++,+++,12761,99,+++++,+++,+++++,+++