Version 2.0.0 beta - Test Candidate 2

This is the second release candidate and fixes several major bugs.

Notable changes:

• Updated striker-installer and InstallManifest.pm to use Red Hat's ' subscription-manager' instead of the now-deprecated RHN.
• Added support in anvil-generate-iso for '--master' to allow ISOs using the latest version from github's HEAD to be used.
• Updated DB->initialize_db() to throw exit code 0 if scancore hasn't run yet on a dashboard so that Gnome's login doesn't fail.
• Updated and fixed several issues related to automatic load shedding.
• Updated and fixed several issues related to automatic live migration, and added a delay to health-based live migrations.
• Updated ScanCore's locking to be more resilient when detecting hung ScanCore/agent instances.

Version 2.0.0 beta - Test Candidate 1

This marks the completion of all features slated for v2.0.0. Striker is now in a feature freeze.

It is fully expected that there will be bugs!

Please open issues with the 'Pre v2.0' tag.

1.2.0 beta TC2

*** NOTE: I'm sticking with the "TC" designation for now, but please note, the changes include enough features and enhancements that it should not have reached "TC" yet. I still don't think it's fair to use the TC label. When a feature-freeze is hit, this notice will be removed. ***

This release includes many enhancements and bug fixes. Below is a summary of the changes from "TC1". These are in chronological order according to commits between TC1 and TC2.

Bugs fixed:

• Fixed a bug (sort of) that prevented a number of VNC ports to open in the firewall to be set in new Install Manifests.
• Cleaned up some mistakes in which strings are presented to users.
• Fixed a path bug in setting permissions in striker-installer.
• Fixed a bug where dhcpd.conf was corrupted when striker-installer was run without defining the BCN.
• Re-enabled sanity checks triggering exit in striker-installer.
• Made it so that striker-installer will abort if either the RHEL or CentOS ISO is defined for the DHCP lease range is not.
• Parsing the DRBD global common and resource files when replacing a node wasn't working, fixed it.
• When replacing node 1, node 2's existing network config wasn't seen triggering a needless remap of node 2, fixed.
• When using spice graphics, the '--graphics spice' wasn't set, fixed.
• Fix a caching problem that prevented Striker from cold-booting an Anvil! that had been built using an install manifest.
• Fixes several bugs when creating only one storage pool.
• Properly maintains the spice/vnc configuration option when the striker.conf is updated.
• fixes a problem where the configuration page's HTML/template was broken when no Anvil! systems were configured yet.
• Fixed uploading files from Striker to a node (expect and checking ~/.ssh/known_hosts were broken).
• Disabled unmounting already-mounted ISOs on Striker when the install is re-run.
• Fixed the backup and restore of Striker config files so that Install Manifests are properly backed up and restored as well.
• Checked for the existence of 'path::power_sh' instead of 'path::inittab' in the striker-installer.
• Fixed Striker installer to only use MTU of 9000 when actually requested.
• Explicitly called 'chkconfig NetworkManager off' because it was enabled on Striker install with '--gui' option.
• Changed the install manifest bash calls that had the string "doesn't exist" to "not found" so that the single quote didn't break the call.
• Updated the generated PXE menu generator to use the proper background image instead of statically setting it.
• Moved the 'manage_daemons()' function in striker-installer down to after the IPMI config and removed the IPMI dependency on NetworkManager (no, I don't know why I did that in the first place, either).
• Fixed a bug where cluster.conf was being clobbered when replacing a node.
• Fixed a bug where a cluster already in the Striker config would not be seen as such, prompting the user to add the repaired Anvil! to the dashboard again.
• Fixed some logging issues.
• Fixed a bug where cman was started on node 1 when it should have been started on node 2.
• Fixed a bug where successfully starting cman was seen as an error and vice-versa.
• Fixed a bug where nodes were always detected as being in a cluster if cman was installed.
• Switched the partition creation code to use GiB instead of MiB as the MiB seemed to be creating unaligned partitions. Will need to enhance this later to not waste space.
• Fixed a bug where setting up DRBD on pool 1 failing would have the failure cleared if pool 2 was ok (or not created). Errors out properly now.

Enhancements Made:

• Added a feature to enable spice graphics in new servers that can be set in the Install Manifest. Deleted the 'switch_spice_to_vnc' system variable in favour of just using the 'use_spice_graphics' system variable.
• Created dedicate kickstart scripts for Striker's PXE server for node 1/2 and striker 1/2. Also made the kickstart scripts for striker machines default selinux to 'permissive'.
• Cleaned up a lot of the output printed by striker-installer.
• Added a pair of new switches to striker-installer to allow for configuring IPMI on striker machines with a BMC.
• Added a new switch to striker-installer for --no-an-repo to prevent (and remove if needed) an.repo from striker machines.
• Started work on adding an MTU variable to the install manifest for pre-configuring jumbo frames.
• Added time-stamps to striker log entries.
• Got the initial version of 'safe_anvil_shutdown' working, including checking the new node health status file.
• When provisioning servers, the bridge was statically set as 'ifn-bridge1'. Changed this to look up the active bridge name.
• Added a check to provision scripts to only call the 'lvremove' when the LV exists, thus avoiding needless error messages on the screen.
• Disabled dhcpd on boot for striker-installer. This is too troublesome and will be addressed soon.
• Got 'striker-installer's "--gui" switch working. This switch now installs a GUI, configures Gnome to shutdown without delay on power button events and configures the user's desktop icons.
• The ability to enable/disable dhcpd (referred to as enabling and disabling the "Install Target") is now added and working.
• Cleans up the configuration page and notifies the user when the global configuration is unchanged.
• Sets SMTP sending and 'helo' domains to use the SMTP user's domain when the domains are 'example.com'.
• The DHCP server can now be toggled on and off via the configuration window.
• Cleaned up the labels/names for the Kickstart files to be more consistent with the rest of the Anvil! naming.
• Cleaned up how fields are auto-filled when creating an Install Manifest with no prefix.
• Fixed the jquery for predicting the IFN IP of dashboards when the IFN is a /24 subnet.
• Finished adding support for custom MTUs being used when set in an Install Manifest.
• Added the button to add the newly built Anvil! to Striker if it wasn't already configured.
• Added support for nodes with name 'node01' and 'node02' instead of just 'n01' and 'n02'.
• Updated the github master URL to point to the direct download file.
• Finished adding support for pre-configuring virt-manager when Striker is installed with --gui.
• Silenced "read-only filesystem" messages when recursively chown'ing /var/www/html with an ISO mounted.
• Cleaned up some text.
• Added a "Cold-Stop Anvil!" button; The entire Anvil! platform can now be taken offline with one button.
• Realized it was 2015 finally...
• Removed 'check_nodes()' function as it was almost literally useless.
• Made the generated kickstart file names shorter and cleaned up the Striker-based generated kickstart entry names.
• Added a check to alert the user if an unaligned partition is created.
• Added a check to see if a node is running cman and, if so, set a flag to block reboots (important if the surviving node has it's kernel updated).
• Added smarts to SSH handling so that, when a node is replaced, stale known_hosts and authorized_keys entries are removed and the new fingerprints/keys added.
• Added some logic so that the warnings about rebooting is not shown when neither node needs a reboot.
• Added some logic so that the password changing warning only printed if a password was about to be changed.
• Made it so that a node's cluster.conf is only written if it's actually changed.
• Installed exists with an error if an unaligned partition is created.

Clean-up and Miscellany:

• The old 'anvil-node-installer' tool has been removed.

1.2.0 beta TC1

*** Test Compose 1 for the v1.2.0 beta release ***

This introduces the new "Install Manifest" feature!

The new "Install Manifest" allows you to answer a few questions, press a button to pre-fill everything else needed by Striker to build an Anvil! system. You review the suggested values, adjust to suite your needs and when you are happy, generate your "Install Manifest". This manifest is stored on the Striker dashboard and can be downloaded for safe keeping or used as a template for another Anvil! system later.

With the manifest, you can point Striker at any pair of nodes with nothing more than a "minimal" RHEL or CentOS 6 install. All you need is each node's current IP address and password. If you are using RHEL proper, you can give it your RHN credentials and it will take care of registration for you, too!

Striker will install all the needed applications, update the OS on the nodes, partition their hard drives, configure the full HA stack, assemble and initialize the clustered storage and leave you with a running, ready to use Anvil!. All you need to do is upload some install media to your shiny new Anvil! and start provisioning your highly-available servers!

Oh, and on that little bit about the nodes needing to have RHEL or CentOS installed?

We've got you covered there, too.

With this new release, your Striker dashboard can be built with the automatic ability to be a PXE boot target with kickstart entries used to build other Striker dashboards and Anvil! nodes.

Someone totally new to HA and/or Linux should be able to have a fully running Anvil! in under a day.

How cool is that? :D

Striker ver. 1.1.6

v1.1.6 - 2014-09-23

• Critical Update!
• Fixed a bug where, if a server's XML definition file is set to use 'spice'
  graphics instead of 'vnc', managing the VM could cause the XML file to be
  blanked out. Recovery is possible by restoring the automatically-generated
  backup, but this is still pretty not-good.
  • This issue has been fixed.
  • Additionally, multiple safe guards have been added to sanity-check the
    definition before writing and after writing, with automatic restore of the
    original XML if the new version appears broken after writing.
• Cleaned up some alert email text.

Striker ver. 1.1.5

NOTE: Please consider a fresh install using the new 'striker-installer' script.

Change log;

v1.1.5 - 2014-08-31

• Fixed some undefined replacement string errors in the server details entry
  for servers not yet added to the Anvil!.
• Disabled the 'Delete' button for a server when it has not yet been added to
  the Anvil!.
• Fixed a variable substitution name in 'message_0211'.
• Made the error displayed when a server fails to be added to the Anvil!
  because of a missing failover domain entry in cluster.conf.
• Added, but commented out, 'no_kill="1"' to new VM services in rgmanager to
  prevent force-off of servers that are slow to boot. This will be enabled as
  soon as resource-agents is updated.
  (as per: https://bugzilla.redhat.com/show_bug.cgi?id=1079032)
• Removed the confirmation requirements for several fairly-safe commands, line
  powering on nodes, booting servers, etc.
• Updated the IPMI scanner code to redirect sensor loss/restore messages to a
  now-smarter debug email address.
• Made all buttons and images come from skins
• Made the Media Library skinable and translatable, finally.
• Added 'id="..."' to all links, input form fields and images.
• Changed 'an-cluster-setup.sh' to no longer disable iptables or selinux.
• Redid the entire main page to return each panel as a completed template chunk
  which is then rendered at once on a final parent template.
• Removed the "Join Anvil!" button's automatic confirmation due to the risk of
  fencing the peer if it is not already in the cluster. Will make this smarter
  later.
• Removed all reference to the dedicated guacamole port (not proxied behind
  apache).
• Explained how to use the on-screen keyboard when displaying a remote desktop
  link and updated the "Trouble?" note to remind the user to make sure no one
  else is connected to a given server already.
• Added a conditional to populating DRBD values when the DRBD daemon is up, as
  it is possible that the daemon can be up before information about a resource
  can be read, leading to "Undefined replacement key" alerts.
• Created the 'striker-install' script to replace the old 'an-cdb-install.sh'
  script.
• Removed all references to SMTP timeout.
• Added a "Back" button to the main Media Library page.
• Reordered the restart calls when restarting guacamole and tomcat, and set it
  up to detect when the start failed and retry up to three times.