FlashArray Collection 1.45.0

v1.45.0

Minor Changes

• Completed the rename to Everpure in the remaining user-visible strings, including the documentation fragment shown by ansible-doc for every module, the FlashArray authentication failure message, and the title of the generated changelog
• api_helpers - Add put_with_context, the verb alias the helper family was missing. PUT endpoints were previously reached through get_with_context.
• module_utils/api_helpers - Add a generic wait_for polling helper, with backoff and timeout, for modules that wait on an asynchronous array operation
• purefa_export - Add server so an export can be created, found and deleted on a file server, as managed by the new purefa_server module. Requires REST API 2.44
• purefa_info - Add C(software) subset, reporting software versions, the patch catalogue and upgrade installation steps.
• purefa_info - Add C(support) subset, reporting support diagnostics and a summary of the system manifest.
• purefa_info - Add logged_in_ports to each host in the hosts subset, mapping every host WWN/IQN/NQN to the array target ports it is currently logged into
• purefa_info - Populate C(performance) and C(vols) for host groups the array does not own. Both loops skipped any name containing a colon, which was how they avoided a KeyError on the host groups the filter had excluded; they now test dict membership, as the C(hosts) and C(pgs) loops already did.
• purefa_info - Report C(is_local) for each fleet member in the C(fleet) subset, identifying which member is the array being queried.
• purefa_info - Report C(is_local) for each host group in the C(hgroups) subset, matching C(hosts), C(pgroups) and C(fleet). A host group in a realm on this array uses a double colon, as C(::), and is reported C(is_local) true, so filtering on the key keeps realm-scoped host groups and drops only those another array owns.
• purefa_info - Report C(is_local) for each host in the C(hosts) subset. On an array in a synchronous replication relationship, C(get_hosts) also returns the peer array's hosts, named C(:). These were indistinguishable from local hosts in the returned dict, so a playbook looping over C(hosts) would pass a remote host name to a module such as C(purefa_host), which rejects it as not conforming to the host naming convention. Filter on C(is_local) to act only on hosts owned by the array being queried.
• purefa_info - Report C(is_local) for each protection group in the C(pgroups) and C(deleted_pgroups) subsets, for the same reason.
• purefa_info - Report per-directory quotas, and the directory user and group quotas and listings, in the C(filesystems) subset.
• purefa_info - Report replication throughput and lag in the C(replication) subset, covering pod replica links, pods, protection groups and snapshot transfers.
• purefa_info - The C(hgroups) subset now reports every host group the array can see, including those it does not own. A host group belonging to another member of the fleet is reported with a single-colon prefix - C(:), or C(:) - and these were previously dropped from the returned dict. A playbook that loops over C(hgroups) and acts on each name will now be handed those names, which modules such as C(purefa_hostgroup) reject as not conforming to the naming convention. Filter on the new C(is_local) key to act only on the host groups the array owns.
• purefa_inventory - Adds slot to each interface in the interfaces subset, reporting the PCI Express slot number that hosts the port.
• purefa_inventory - Hardware components without a dedicated dict of their own are now reported in a new other dict, keyed by component name and carrying the component type, rather than being dropped from the inventory. This covers direct_compress_accelerator, sas_module and storage_shelf, as well as any component type the array reports in future.
• purefa_inventory - Reports nvram_bay hardware components in the drives subset, so NVRAM devices carry identify_enabled and serial like every other drive. nvram_bays are already captured in this dict by the array.get_drives() pass but would otherwise be missing the identify_enabled key.
• purefa_inventory - Temperature sensors are now reported in the temperature dict, which was previously initialized but always empty. They continue to be reported in controllers as well, so this is a backwards compatible addition.
• purefa_inventory - The contents of the other dict are not stable. A component reported there may be moved into a dedicated dict in a future release, so match on the type key rather than assuming a component stays in other.
• purefa_user - Document that setting public_key is not idempotent. The array does not expose the actual public key value, so it cannot be compared against the provided key so a task that sets one always reports as changed, same as passwords. Removing a key is idempotent.
• purefa_volume_tags - Warn that the module is superseded by M(everpure.flasharray.purefa_tags), which manages volume tags as well as tags on every other taggable resource.
• purefa_workload - A task naming neither context nor placement now fails, naming the members involved, when two or more fleet members hold the workload name. In 1.44.0 it acted on the copy held by the array receiving the request, which is not necessarily the one intended. A playbook that relied on that behaviour needs an explicit context or placement.
• purefa_workload - Add wait and wait_timeout so a task waits for the array to finish before returning, on by default and required whenever host is set
• purefa_workload - Name the fleet member a context-less delete or eradicate resolved to, before anything is removed from it
• purefa_workload - Refuse an explicitly empty context rather than reading it as unset, so that a value computed from an undefined variable cannot silently widen a destructive task to the whole fleet
• purefa_workload - Return a workload dict on every action describing the resulting fleet member, status, and volumes, instead of only changed
• purefa_workload - Return the workload name and fleet member from a delete or an eradicate, in check mode and on a real run alike. An eradication leaves nothing else to read, so those two keys are all it reports
• purefa_workload - Warn on every action, not only the destructive ones, when a workload name already exists on another fleet member, since a name is unique per member, not per fleet. A rename asks for both names, since renaming onto a name another member already holds produces two of it
• purefa_workload - context/placement now default to the whole fleet rather than to the array receiving the request, which is also what naming the fleet itself means: "the workload with this name, wherever it is in the fleet". Only works for a create when it is used together with recommendation.
• purefa_workload - eradicate: true is now refused on any state but absent, and host is refused together with rename. Both were previously accepted and ignored, so a playbook setting either will now fail rather than silently doing nothing with it.

Deprecated Features

• purefa_inventory - Reporting temperature sensors in the controllers dict is deprecated and will be removed in the next major release. Use the temperature dict instead, so that controllers holds only controllers.

Bugfixes

• purefa_audits - Fixed the pytz requirement check never firing, so a missing pytz failed with NameError instead of reporting the missing requirement
• purefa_directory - Fix rename so that re-running a rename task reports no change, instead of failing with a path collision or creating a duplicate directory under the original name.
• purefa_directory - Report a failure when the parent file system is destroyed, rather than silently reporting no change for a rename that the array would refuse.
• purefa_eula - Fixed the already-signed check, which used hasattr(current_eula, "signature.accepted"). hasattr does not traverse attributes, so the test was always false and the EULA was re-signed on every run, always reporting changed. It now checks signature.accepted, which the array returns as null until the EULA is accepted
• purefa_fs - Fix rename so that re-running a rename task reports no change, instead of recreating the file system under its original name.
• purefa_host - Reconcile a host's WWNs, IQNs and NQNs by adding only missing initiators and removing only unwanted ones, so one can be dropped by omitting it and those that stay are untouched
• purefa_info - Added missing total_effective to destroyed volumes, which is already returned for live volumes
• purefa_info - Fix TypeError gathering the hosts subset for hosts that have preferred arrays set, as preferred_arrays is a list of references not a dict
• purefa_info - Fix destroyed volume snapshots always being reported with C(is_local) set to true in the C(deleted_snapshots) subset. A snapshot replicated from another array is now correctly reported as not local, matching the behaviour of the C(snapshots) subset.
• purefa_info - Fix the C(admins) subset, and therefore C(all), crashing on an array whose administrators have no public key. The field is null by default and the SDK raises rather than returning None.
• purefa_info - Fix the C(hgroups) subset raising a KeyError when a host group the array does not own carries a tag. The tag loop indexed the returned dict by resource name without a guard, but the host group itself had been filtered out of that dict.
• purefa_info - Fixed AttributeError traceback when an array returns a null space value, which aborted the whole gather. The SDK client is created with model_attribute_error_on_none at its default of True, so reading a declared-but-null field raises instead of returning None. All space attribute reads are now guarded, affecting the volumes, snapshots, pods, vgroups, filesystems and capacity subsets
• purefa_info - Fixed data_reduction, total_reduction and used_provisioned being returned as single-element lists instead of numbers, caused by a trailing comma wrapping the value in a tuple, for volumes, volume groups and volume snapshots
• purefa_inventory - Fixes the serial and identify_enabled of each drive being dropped from the drives subset, as the array.get_drives() pass overwrote the value already collected from the matching drive_bay hardware component.
• purefa_network - Fixed AttributeError reading the MTU or the service list of an interface the array reports as null
• purefa_network - Fixed AttributeError setting an address on an interface that has no gateway configured
• purefa_network - Fixed AttributeError updating an interface that has no subinterfaces, such as any plain physical port
• purefa_network - Fixed AttributeError when setting an address without a gateway on an interface that has no gateway configured
• purefa_network - Fixed a compatible existing gateway being rejected with "Gateway and subnet are not compatible" when changing the address of an interface without also supplying a gateway
• purefa_network - Fixed removing an address or gateway from an ethernet interface always reporting changed
• purefa_pg - Fix a protection group in a pod that is itself in a realm, named C(realm::pod::pgroup), being treated as though the realm were the pod. Containers nest, so the pod is C(realm::pod), but the pod existence check looked for a pod named after the realm alone and failed with "Pod does not exist", and the naming convention check validated the pod name rather than the protection group name.
• purefa_pgsnap - Fix C(restore) from a protection group in a pod that is itself in a realm. The volume to restore was qualified with the realm alone, as C(realm::volume), rather than with the whole container, so it never matched a volume in the protection group.
• purefa_policy - Fix AttributeError updating a password policy when the array returns null for fields such as lockout_duration, by reading them safely and treating null as not configured
• purefa_policy - Fix rename so that re-running a rename task reports no change, instead of recreating the policy under its original name.
• purefa_policy - Fix rename to report a change in check mode, rather than always reporting no change.
• purefa_policy - Gate max_password_age behind FlashArray REST API 2.39, the version that introduced it, so a clear error is returned instead of an unsupported-field failure on older arrays
• purefa_sessions - Fixed the pytz requirement check never firing, so a missing pytz failed with NameError instead of reporting the missing requirement
• purefa_smtp - Fixed encryption_mode not being cleared when set to an empty string
• purefa_smtp - Fixed state=absent always reporting changed even when the SMTP settings were already cleared
• purefa_smtp - Fixed the module always reporting changed when any SMTP setting is left unconfigured on the array
• purefa_user - Fixed check_mode making real changes for an AD user, where a run would delete and recreate the API token and change the SSH key despite the module declaring check mode support. The same issue was present for both the API token and the SSH key of an existing local user.
• purefa_user - Fixed check_mode reporting no change from state: absent for an AD user that has array-side state to remove. changed was assigned inside the check mode guard, so a --check run reported nothing while a real run reported a change.
• purefa_user - Fixed public_key: "" being silently ignored for an AD user, so the documented way to remove a key now works. Uses the same logic as the local user task.
• purefa_user - Fixed an AD user being rejected when their name does not meet Purity's local username rules, such as the first.last form used by a directory service. A name is now checked against the local pattern only when ad_user is false. An AD name may use any script, so that a directory is free to name its own users, and is rejected only for the characters that would change which admins the request targets - whitespace, C0 and C1 control characters, and any of ,, /, ?, &, # - or for being empty or over 128 characters.
• purefa_volume - Fix the same mishandling of a C(realm::pod::pgroup) value passed to the C(pgroup) option, where the naming convention check validated the pod name rather than the protection group name, and the pod existence failure named the realm alone.
• purefa_volume_tags - Accept C(keys) as an alias for I(tag), matching the option name used by M(everpure.flasharray.purefa_tags).
• purefa_volume_tags - Fix I(state=absent) so that it actually removes tags. The keys from I(kvp) were compared as single-element tuples against key strings, so nothing ever matched and the module reported no change.
• purefa_volume_tags - Fix the I(tag) option, which was documented and accepted but never read, so removing tags by key did nothing. Used on its own it also raised a TypeError.
• purefa_workload - Connect host on a re-run of a recommendation-based create, which previously returned before the host was connected
• purefa_workload - Connect or disconnect host only where needed, so naming a host already connected to some of the workload's volumes no longer fails
• purefa_workload - Fail on an array error during the workload lookup instead of reading any non-200 response as the workload being absent, which could let a create duplicate an existing workload
• purefa_workload - Fail with a clear message instead of a traceback or crash when a workload cannot be found, its volumes cannot be listed, or Fusion cannot find a placement
• purefa_workload - Fail, rather than reporting no change having added nothing, when state=expand targets a destroyed workload
• purefa_workload - Fail, rather than reporting no change, when state=expand targets a workload that does not exist
• purefa_workload - Fixed a bug which caused imposible renames to create or recover a workload
• purefa_workload - Fixed fleet-wide idempotency for create, delete, and eradicate, which previously only checked the named member rather than the whole fleet
• purefa_workload - Fixed idempotency with recommendation, re-running found a new placement each time and created a duplicate workload instead of recognising the one already there
• purefa_workload - Recover a since-destroyed workload found by a recommendation-based create, instead of reporting no change
• purefa_workload - Refuse rename together with state=absent or state=expand, rather than silently ignoring it
• purefa_workload - Reject volume_count: 0 during an expand operation. This is a bugfix for a regression introduced in 1.0.34.
• purefa_workload - Require preset for state=expand, and fail with a message naming the option, rather than reporting that preset None does not exist
• purefa_workload - Restrict the fleet member list to the fleet in use, so the two halves of a fleet-wide search cannot look at different sets of arrays on an array belonging to more than one fleet
• purefa_workload - Stop state=absent with both host and eradicate: true from eradicating a destroyed workload. Naming a host makes the task a disconnect, and the volumes went with the workload, so the task now reports no change with a warning
• purefa_workload - Stop state=expand from creating volumes for real while running in check mode
• purefa_workload - Stop a create using recommendation from failing with an internal error when no context is named
• purefa_workload - Warn, rather than saying nothing, when volume_count, volume_configuration or parameters is set on a task that does not read it.
• purefa_workload - placement now takes effect on every action; it was previously ignored outside of a recommendation-based create.
• purefa_workload - preset is only read when creating or expanding a workload, so omitting it on a delete, rename, recover, or host operation no longer raises a TypeError

New Modules

• everpure.flasharray.purefa_lds - Manage Everpure FlashArray local directory services
• everpure.flasharray.purefa_localgroup - Manage Everpure FlashArray local groups
• everpure.flasharray.purefa_localuser - Manage Everpure FlashArray local users
• everpure.flasharray.purefa_server - Manage Everpure FlashArray file servers
• everpure.flasharray.purefa_tags - Manage Everpure FlashArray resource tags