* Fixed a bug in Cluster->parse_cib() where the local machine's ready state was being set to the node name.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated DRBD->allow_two_primaries() to take the 'set_to' parameter which can be 'yes' to all and 'no' to disallow dual-primary.
* Updated ocf:alteeve:server to call allow_two_primaries() with 'set_to' = 'no' instead of calling 'adjust' after a migration completes.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated scan-cluster to check / set which node should be preferred if a netsplit causes a fence race.
* Fixed a bug in Server->shutdown_virsh() where a shutdown timeout would go into a loop.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Database->archive_table() and ->_find_behind_databases() to loop through connected databases, instead of configured databases.
* Updated Network->get_ips() to only record the real MAC addresses on network interfaces (not bonds or bridges) in the "network::${host}::interface::${in_iface}::mac_address" hash. This should help avoid reboot loops caused by anvil-configure-host thinking the network needs to be reconfigured when it doesn't actually need to be.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Cluster->parse_cib() to not require a database connection (part of the work to make ocf:alteeve:server run without a DB)
* WIP: Continuing work on the ocf:alteeve:server RA to run without database connections.
* Updated the scancore daemon to explcitely check that all scan agent schemas are loaded in all databases on startup. This is to resolve resync issues on rebuilt strikers that may not yet have some schemas loaded when a DB resync runs.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Changed the alteeve repo RPM to the new cimmunity/enterprise repo
* Fixed a bug where 'fence_data::updated' was causing the fences web page to break.
* Fixed a bug in Database->insert_or_update_network_interfaces() where certain interfaces were being repeatedly added to the database.
* Fixed a bug in Database->_find_behind_databases() was marking DBs as behind even though they had less than 10 columns off.
* Fixed a bug in Get->host_name() where, if the host name was changed on disk but the environment variable was still the old name, it would cause the hostname to waffle back and forth and cause constant updated to /etc/hosts.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created Cluster->check_stonith_config() that checks and, if needed, reconfigures a cluster's fencing (stonith) config.
* Updated scan-cluster to call Cluster->check_stonith_config() at the end of each call.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created Cluster->get_fence_methods() that parses all fence methods out of a recorded CIB and stores the in a hash for a given host_uuid.
* Fixed a bug in ScanCore->post_scan_analysis_striker() where the short_host_name was not being stored correctly.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created DRBD->check_if_syncsource() and ->check_if_synctarget() that return '1' if the target host is currently SyncSource or SyncTarget for any resource, respectively.
* Updated DRBD->update_global_common() to return the unified-format diff if any changes were made to global-common.conf.
* Created ScanCore->check_health() that returns the health score for a host. Created ->count_servers() that returns the number of servers on a host, how much RAM is used by those servers and, if available, the estimated migration time of the servers. Updated ->check_temperature() to set/clear/return the time that a host has been in a warning or critical temperature state.
* Finished ScanCore->post_scan_analysis_node()!!! It certainly has bugs, and much testing is needed, but the logic is all in place! Oh what a slog that was... It should be far more intelligent than M2 though, once flushed out and tested.
* Created Server->active_migrations() that returns '1' if any servers are in a migration on an Anvil! system. Updated ->migrate_virsh() to record how long a migration took in the "server::migration_duration" variable, which is averaged by ScanCore->count_servers() to estimate migration times.
* Updated scan-drbd to check/update the global-common.conf file's config at the end of a scan.
* Updated ScanCore itself to not scan when in maintenance mode. Also updated it to call 'anvil-safe-start' when ScanCore starts, so long as it is within ten minutes of the host booting.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Fixed a bug where, in rare cases, $anvil->hostname() would call 'hostnamectl' and get a dbus error during shutdown, which would then cause the hostname to be changed to the error in the database.
* Fixed a bug in Cluster->boot_server() where it would never verify that a server has started successfully.
* Updated Database->get_ip_addresses() to store the IPs we manage in 'ip_addresses::<ip_address_address>::X'.
* Updated ocf:alteeve:server to work from command line calls, though more testing is still needed.
* Started work on 'anvil-rename-server', but haven't gotten far with it yet.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Turned on print to console for logging in anvil-provision-server. Also updated it to check if the cluster is running and hold until it is.
* Cleaned up some code in Get->available_resources() that proved hard to debug.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created Cluster->parse_quorum() to check if a node is quorate as 'have-quorum' in the pacemaker CIB doesn't appear to be super accurate during startup.
* Fixed a bug in striker-manage-install-target where if a node didn't have any registered IPs, it would break before generating the repo data.
* Fixed a bug in anvil-join-anvil where if the database had to be reconnected, the job data was lost.
Signed-off-by: Digimer <digimer@alteeve.ca>
* The get_mmeory endpoint was completed.
* The get_replicated_storage endpoint was completed, though it requires testing and likely has issues.
To prepare for the get_status endpoint work, I needed to update ScanCore and modules to track the host_status. This commit contains the work needed for this.
* Updated ScanCore->post_scan_analysis_striker() to use configured fence devices (except PDUs) to check if a target host is off or on, in there is no host_ipmi interface. In all cases, if a machine can be confirmed on or off, the host_status is now updated.
* To support the above fence based power checks, updated scan-cluster to store the on-disk CIB in the new scan_cluster -> scan_cluster_cib colume.
* Updated ScanCore->parse_cib() to map stonith primitive IDs to fence agents. Updated ->parse_crm_mon() to not call if the executable doesn't exist to avoid unhelpful error messages in the logs when called from a Striker.
* Update DRBD->gather_data() to get the size data from /sys/block/drbd<minor>/size' x '/sys/block/drbd<minor>/queue/logical_block_size so it works when a device is Secondary (and can't be promoted).
* Updated Database->get_hosts_info() to record the short host name as well as the stored host name. Created ->update_host_status() as a wrapper to ->insert_or_update_hosts() that only updates the host status.
* Updated anvil-join-anvil to disabled ksm and ksmtuned daemons.
* Updated scancore and anvil-daemon to set the host_status to 'online' on startup.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Database->insert_or_update_storage_group_members() to use the host_uuid when trying to find existing members.
* Added the skeleton of a bunch of new json endpoints for the new UI features.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Reworked DRBD->get_next_resource() to pull from the database, and to no longer do that increments-of-three nonsense. Avoidable complexity. Also added a call to Cluster->get_anvil_uuid() if the 'anvil_uuid' parameter wasn't passed.
* Updated Database->get_host_from_uuid() and ->get_hosts() to now take 'include_deleted' parameter and default to not returning deleted hosts. This fixed issues where anvil-{delete,provision}-server calls could assign jobs to now-deleted hosts with reused host names.
* Updated anvil-delete-server to print log entries to STDOUT. Also updated it to not wait of shutdown of a server in pacemaker to complete, and instead to destroy it after calling pacemaker's resource stop. Updated to also check to see if the server being deleted is already out of pacemaker and, if so, skip that step and directly try to destroy the server, if it's running.
* Updated anvil-provision-server to force 'peer_mode' runs to pull their TCP Port and DRBD minor numbers from the job. This fixes a bug where the same resource on two machines could use different TCP ports.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Reworked DRBD->get_next_resource() to pull from the database, and to no longer do that increments-of-three nonsense. Avoidable complexity. Also added a call to Cluster->get_anvil_uuid() if the 'anvil_uuid' parameter wasn't passed.
* Updated Database->get_host_from_uuid() and ->get_hosts() to now take 'include_deleted' parameter and default to not returning deleted hosts. This fixed issues where anvil-{delete,provision}-server calls could assign jobs to now-deleted hosts with reused host names.
* Updated anvil-delete-server to print log entries to STDOUT. Also updated it to not wait of shutdown of a server in pacemaker to complete, and instead to destroy it after calling pacemaker's resource stop. Updated to also check to see if the server being deleted is already out of pacemaker and, if so, skip that step and directly try to destroy the server, if it's running.
* Updated anvil-provision-server to force 'peer_mode' runs to pull their TCP Port and DRBD minor numbers from the job. This fixes a bug where the same resource on two machines could use different TCP ports.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Cluster->add_server() to now set failure timeouts to actual numbers instead of INFINITY after discovering that INFINITY doesn't work in those cases.
* Updated Databsae->get_hosts to now check if other entries have the same host name, and if so, to set their host_key to 'DELETED'. This should make it easier to handle when a hardware machine is replaced by new hardware but uses the same host_name.
* Updated Email->check_queue() to start and enable postfix.service if it's found to not be running.
* Updated Get->available_resources() to return '!!no_data!!' when a given host hasn't got any data in scan_lvm_vgs. Now use this in anvil-provision-server to exit if a node or dr host hasn't run scancore yet.
* Fixed a bug in scan-lvm where the pvs_uuid wasn't being loaded properly, preventing lost PVs, VGs and LVs from being flagged as deleted.
* Started work on anvil-migate-server, though it's far from complete.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Fixed a bug in Cluster->shutdown_server() where the wrong variable was being evaluated when checking the server state.
* Created DRBD->delete_resource() that deletes a resource's backing device and configuration. Note that this wipes the DRBD MD and and FS signatures before removing the LV. Updated DRBD->gather_data() to record the backing devices for volumes.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Fixed a bug in Cluster->parse_cib() when a server that is off wasn't setting 'status'.
* Renamed 'server::location::<server>::host' to '...::host_name' in several places.
* Got more work done on anvil-delete-server, up to the point where it calls the new Cluster->delete_server() method.
* Updated fence_pacemaker to call 'drbdadm adjust all' to dampen an issue where in-memory fence configs seem to change, preventing reconnection of the peer after it reboots from the fence. More testing needed on this issue.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Removed the exit-if-no-DB check in ocf:alteeve:server so that (hopefully, needs testing), running servers won't be impacted if the nodes lost contact with both/all strikers.
* Updated scan-server to make an explicit check for missing XML definition files on startup and write them if needed.
* Very beginning work on anvil-delete-server has been started.
* Updated anvil-provision-server to wait when it's running in peer mode until the new XML definition is in the DB and then write it out to disk before exiting. Also updated it to add the new server to pacemaker before exiting.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created the new method Cluster->get_primary_host_uuid() that returns the 'host_uuid' of the primary node in the given cluster. This is useful for external programs to figure out which node is primary. Example is provisioning a new server being assigned to the active node. Also created ->is_primary() that is a similar test to see if the active node is the primary node or not.
* Updated Cluster->parse_cib() and ->parse_crm_mon() to work on remote hosts.
* Updated Database->get_hosts() to store the short host names.
* Created Get->host_from_ip_address() that translates an IP address to a host_uuid and host_name, if it's an IP assigned currently to a known host.
* Created Network->find_target_ip() that simplifies finding which IP address to use when the caller wants to connect to a target host.
* Reworked the anvil-join-anvil to parse fence_arguments in a way that handles passwords with spaces in them.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created Cluster->assemble_storage_groups() and moved the logic to auto-assemble groups out of Get->available_resources().
* Created Cluster->get_anvil_name() that will return an Anvil! name for a given anvil_uuid, or the name of the Anvil! if the host is a member of an Anvil!.
* Updated Cluster->get_anvil_uuid() to return the 'anvil_uuid' if passed a specific 'anvil_name'.
* Updated Jobs->clear() to use 'switches::job-uuid' when a job_uuid is not passed but the value exists in 'switches::job-uuid'.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created (but not finished) scan-apc-pdu
* Added support to tracking maintenance-mode for nodes in Cluster->parse_cib
* Created Remote->read_snmp_oid().
* Created Server->get_definition.
* Updated Server->get_status() to write-out server XML files on-demand.
* Finished scan-cluster.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updates servers -> server_host_uuid to drop the 'NOT NULL' constraint.
* Created the new Get->server_uuid_from_name() that does what it says on the tin. Fixed a bug in ->host_uuid_from_name() where the host name was being returned instead of the UUID.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Renamed servers -> 'server_clean_stop' to 'server_user_stop' to make it clearer what the column represents.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Server->migrate_virsh() to set 'servers' -> 'server_state' to 'migrating' and clear it again once the migation completes. Also added support for cold (frozen) versus live migrations.
* Updated Cluster->parse_cib() to check if a server with the server_state set to 'migrating' isn't actually migrating anymore and, if not, to clear that state. This is needed as scan-server will blindly ignore/skip any migrating server, and if a migration call is interrupted, the state could get stuck.
* Updated the 'servers' database table (and associated Database methods) to add columns for;
** server_ram_in_use - tracking RAM used by a running server
** server_configured_ram - RAM allocated to a running server (used with the above to alert a user and track _currently_ available RAM)
** server_updated_by_user - To be set by Striker tools to indicate when the user made a change that needs to push out to nodes / running server.
** server_boot_time - Tracks the unixtime when the server booted (to track uptime even if the server migrates across nodes).
* Created Get->anvil_name_from_uuid() to easily convert an Anvil! UUID into a name. Also created ->host_uuid_from_name() to translate a host name into a host UUID.
* Created Server->get_runtime() that translates a server name into a process ID and then uses that to determine how long (in seconds) it has been running. This is used when a server transitions from 'shut off' to 'running' to determine exactly when the server booted (current time - runtime).
* Renamed all 'Server->parse_definition' calls that used 'from_memory' to 'from_virsh' to clarify the data source.
* Made scan-hardware smarter about RAM change alerts.
* Updated scancore to load agent strings on startup so that processing pending alerts works properly.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Renamed the 'defitintions' table to 'server_definitions' to clarify the purpose, and made all the 'server' columns have then 'not null' constraint.
* Created Database->insert_or_update_servers(), ->get_servers(), ->insert_or_update_server_definitions() and ->get_server_definitions().
* Updated scancore, anvil-daemon, and scan agents to not run unless they're run with root privs.
* Got scan-server to update the servers / server_definition tables and the on-disk file when needed.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Began (but haven't finished) Database->insert_or_update_servers().
* Created Storage->get_file_stats() to collect the (l)stat information for a file.
* Got more work done on scan-server.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Added support in anvil.conf to disable scan agents with 'scancore::<agent_name>::disable', and added handling this to agents. Also allowed for '--force' to override this setting.
* Updated ScanCore->agent_startup() to allow for empty scan agent table lists.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Did more work on parsing server data out of the CIB. There is still an issue with determining which node currently hosts a resource, however.
* Renamed Server->boot to ->boot_virsh, ->shutdown to ->shutdown_virsh and ->migrate to ->migrate_virsh to clarify that these methods work on the raw virsh calls, outside of pacemaker (indeed, they are what the pacemaker RA uses to do what pacemaker asks).
* Got more work done on the scan-cluster SA.
* Created the empty files for the pending scan-server SA.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created 'Cluster->which_node' that returns 'node1' or 'node2' to indicate which node a host is.
* Continued working on scan_cluster; decided to make it not host-dependent.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Converted most 'eval { }' calls to localize $@ and test the output of the eval, instead of checking to see if $@ was set.
* Converted all 'local' hash references to instead use the short host name of the local machine as a new standard.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Cleaned up a lot of logging.
* Updated Cluster->parse_cib() to track if a stonith device has 'delay' set.
* Got a lot more work done on anvil-join-anvil's stonith processing, but it still isn't complete. Updated it to change shell user passwords as well.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Cluster->parse_cib() to take a CIB as a parameter.
* Fixed a bug in Database->get_hosts() where loading the host_ipmi value was filtered through Log->is_secure.
* Updated Striker->get_fence_data() to parse the switches to make it easier to map a fence agent's command line switches to STDIN arguments.
* Created System->parse_arguments() that converts a series of command line switches and their values into a hash. It's similar to Get->switches(), but works on any string.
* Continued work on anvil-join-anvil's fence configuration logic.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Updated Storage->rsync() to only create the rsync wrapper if a password was given, allowing for rsync to work to/from a remote system when passwordless SSH is enabled.
* Updated anvil-join-anvil to disable/stop drbd.service, and to properly wait until both nodes are online.
Signed-off-by: Digimer <digimer@alteeve.ca>
* Created Cluster->get_peers() that figures out who the peer node (and DR host, if applicable) are.
* Updated Cluster->parse_cib() to dig out more information.
* Created Cluster->start_cluster() to start pacemaker (via pcsd) locally or on all (both) nodes.
* Started working on ocf:alteeve:server to start/stop the libvirtd/drbd daemons as needed, instead of having pacemaker do it.
* Got more work done on anvil-join-anvil. Node 2 now waits for the cluster to start, and node 1 will do setup as needed, then wait for the cluster to start as well.
Signed-off-by: Digimer <digimer@alteeve.ca>
Digimer