You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
v1.46.0
Minor Changes
CI - The artifact job now fails on any symlink in the build, on unreleased fragments or test directories being shipped, and when CHANGELOG.rst has no section for the version in galaxy.yml. It also runs ansible-lint against the artifact, not only ansible-test sanity.
README - Converted the badge block from unbalanced raw HTML to markdown image links, and reconciled the Available Modules list against the collection - it listed 73 of 75 modules, misspelled purefa_audits, and omitted purefa_dsrole_old and purefa_kmip.
all - the minimum py-pure-client version is raised from 1.82.0 to 1.88.0, the first release that carries the REST 2.54 models needed by the new purefa_preset module. The everpure.fa documentation fragment, which had said 1.90.0 since the 1.44.0 rebrand without anything needing it, now gives the same 1.88.0 as requirements.txt and the README.
collection - Narrowed the changelogs exclusion so that changelogs/changelog.yaml still ships alongside the rendered CHANGELOG.rst while antsibull-changelog tooling state, its configuration, and unreleased fragments do not.
collection - Removed the playbooks/roles symlink and stopped shipping the empty scaffolding directories. A symlink is recorded in FILES.json as an ordinary directory with a null checksum, so nothing behind it is verified, and content scanners routinely skip symlinks.
collection - tests/README.md no longer ships. It documents the tests/unit directory, which is excluded from the build, so none of the paths it describes exist in the collection.
purefa_apiclient - token_ttl is documented as seconds (1-86400), converted to milliseconds for the array; 86400 is applied only when creating a client, so an unset token_ttl is not compared against an existing client.
purefa_audits - the purefa_audits return value is documented; a dictionary keyed by audit record name with time, arguments, user, command, subcommand and origin for each event.
purefa_dns - the service option no longer defaults to management. Omitting it creates a DNS configuration with no service, which is what every file server after the first uses through the dns option of purefa_server. A play that gave a custom name without service used to have the name rewritten to management; it now creates a configuration with that name, so add service=management or use the management name directly where the management DNS was meant. The limit of two configurations is removed; the array itself enforces one configuration per service.
purefa_eradication, purefa_host, purefa_policy, purefa_sessions, purefa_network_access_policy - document array rules learnt on hardware (disabled_delay cannot be decreased, Pure OUI WWNs are refused, SMB rules share access and encryption settings, unenforced quota limits, sessions without a start time, positional rule names).
purefa_hardware - enabled is now required; an unknown component fails with a clear message.
purefa_info - Add servers gather_subset to report file server configuration
purefa_info - Network-access policies in the policies subset now report their real rules. The rules field was previously hardcoded to an empty list for every policy of this type, even when the policy had real rules configured.
purefa_info - QoS policies in the policies subset now report max_total_bytes_per_sec, max_total_ops_per_sec, pod, and (REST 2.55+) realms.
purefa_info - TLS policies in the policies subset now report appliance_certificate, client_certificates_required, disabled_tls_ciphers, enabled_tls_ciphers, min_tls_version, tls_enforced_for, trusted_client_certificate_authority, and verify_client_certificate_trust.
purefa_info - report attached_server for each Ethernet interface in the network subset, naming the file server the interface serves data for.
purefa_info - report the login banner as default.banner, alert watchers as config.alert_watchers, directory exports, every SAML identity provider under config.saml2sso keyed by name, management-access policy rules and aggregation strategy, and extra password and SMB policy fields.
purefa_info - the dns section of the config subset reports name and services for each configuration, and lists a configuration with no service under its name.
purefa_network - add server option to attach an Ethernet interface to a file server when it is created, rather than having to move it afterwards with purefa_server. Omitting the option keeps the existing behaviour, in which the array attaches a new interface to its default _array_server. Set it to an empty string to attach an interface to no file server at all. Requires Purity//FA REST API 2.44 or higher.
purefa_network - virtual interface names containing hyphens are rejected before the array is called.
purefa_pg - a replication target can be removed with state absent and target.
purefa_pg, purefa_pgsched, purefa_vg, purefa_eradication, purefa_policy_management_access - enabled, priority_operator/priority_value, the eradication delays and enabled/aggregation_strategy are only applied when supplied instead of being reset to an implicit default on every update.
purefa_policy_management_access - rule roles are validated against the four the array accepts (viewer, support, admin, storage).
purefa_snap - add with_default_protection for volumes created by a snapshot copy.
purefa_user - a new user still gets readonly when neither role nor access_policy is set, but an existing user's role or policy is now only changed when one of them is given. Before, omitting role reset an existing user to readonly.
purefa_user - add access_policy to make a local user a member of a custom management access policy instead of giving it a built-in role (requires REST 2.36). The user is then limited to what the policy grants, for example a user scoped to one realm for Portworx/PX-CSI secure multi-tenancy. role and access_policy are mutually exclusive.
purefa_volume - the documentation shows that a volume in a pod that lives in a realm is addressed as REALM_NAME::POD_NAME::VOLUME_NAME.
Bugfixes
purefa_ad, purefa_offload - version checks use comparisons instead of substring tests; purefa_offload azure account handling.
purefa_admin - lockout 0 clears the lockout and sso false disables single sign-on.
purefa_apiclient - a run whose public_key, issuer, role or token_ttl differ from an existing API client warns instead of silently reporting ok; the array only allows enabled to be changed in place, so the client must be deleted and recreated to apply them.
purefa_audits - an audit record that does not carry every field (for example arguments, subcommand or origin) no longer raises AttributeError; the missing field is returned as none.
purefa_audits - the start and timezone option descriptions are corrected; timezone is the zone in which start is interpreted and the returned times are reported, and the array's own timezone is only read when timezone is omitted.
purefa_cbsexpand, purefa_certs, purefa_connect, purefa_console, purefa_dns, purefa_ds, purefa_dsrole, purefa_dsrole_old, purefa_endpoint, purefa_eula, purefa_fleet, purefa_fs, purefa_hg, purefa_host, purefa_kmip, purefa_maintenance, purefa_ntp, purefa_pg, purefa_pgsched, purefa_pgsnap, purefa_phonehome, purefa_pod, purefa_proxy, purefa_ra, purefa_realm, purefa_smtp, purefa_snmp, purefa_syslog, purefa_syslog_settings, purefa_tgroup, purefa_user, purefa_vg, purefa_vnc, purefa_volume, purefa_workload - array error messages no longer double the word failed (for example Delete pod x failed failed.).
purefa_certs - generate true regenerates the key and days is applied.
purefa_connect - an existing connection is matched by address and type, the fleet-management connection is never touched, and encrypted uses the correct version check.
purefa_default_protection - pod-scope add and remove send the correct body.
purefa_dirsnap - keep_for is compared before it is patched; recovery no longer sends a null retention.
purefa_dns - deleting a configuration name that does not exist is no longer redirected to the management configuration, which emptied the management DNS.
purefa_dns - emptying the management configuration now honours check mode.
purefa_dns - update the management configuration when it carries several services, add a source to a configuration without one, only send source when given, and never redirect state absent of an unknown name to the management configuration.
purefa_dns - updating a configuration failed with "Changing service type is not permitted" when it carried more than one service, as the management configuration does on arrays where it also serves files.
purefa_ds, purefa_dsrole - a repeated state absent is a no-op when the array returns empty values.
purefa_endpoint - destroy and eradicate are idempotent and reachable; a rename repeat no longer re-creates the endpoint.
purefa_eradication - enabled_delay and disabled_delay no longer default to 1 day. A delay that is not given keeps the array's current value and a call with neither is a no-op; previously changing one delay silently reset the other to one day. timer now sets both delays when the array no longer reports eradication_delay because they differ.
purefa_fleet - member removal no longer raises; delete honours check mode.
purefa_fs - moving a file system between pods reads the source pod.
purefa_hardware - the identify LED request was built with the Hardware response model, which serialises to an empty body, so the array answered 400 "No operation was specified" and the LED never changed. The request is now built with HardwarePatch.
purefa_hardware - wait for the identify LED state to settle after a change so an immediate repeat is a no-op.
purefa_hg - a repeated rename is a no-op; deleting an empty host group reports a change; lun without volume fails cleanly.
purefa_hg - deleting a host group with more than one member host and connected volumes no longer fails with Unidentified internal error; each volume is disconnected once instead of once per host.
purefa_host - preferred_array error message, count with state absent fails clearly, CHAP clearing is idempotent.
purefa_info - A protection group snapshot with no transfer record is now reported. Only a replicated snapshot has one, so a protection group with no replication target previously reported no snapshots at all.
purefa_info - Fixed created and time_remaining being null for every protection group snapshot reported under the pgroups and deleted_pgroups subsets. Both were hardcoded to None when the module moved to REST 2 in 1.37.0, because they are fields of the snapshot and the module only read the snapshot's transfer record, which carries neither at any REST version.
purefa_info - destroyed is no longer read from a protection group snapshot transfer record without a default, which would raise AttributeError on a record that omits it.
purefa_info - gather_subset all no longer crashes in the host-group performance dict or drops the default section; performance keys read the attribute of the same name; replica links report recovery_point.
purefa_info - offload-target snapshots of pod and realm volumes are keyed by their full pod::vol.suffix or realm::pod::vol.suffix name in the snapshots and deleted_snapshots sections, and their transfer details are looked up by name rather than by the snapshot object.
purefa_info - the config subset failed with an IndexError when the array had a DNS configuration with no service.
purefa_info - the fleet, presets and workloads sections are empty instead of a traceback when the array does not support Fusion or the read does not succeed.
purefa_info - the presets and workloads subsets are gathered from REST 2.40, where py-pure-client provides the calls, instead of failing with AttributeError on REST 2.38 and 2.39 arrays; fleet is still gathered from REST 2.38.
purefa_inventory - ports missing from the hardware listing no longer raise.
purefa_kmip - the CA certificate length is validated instead of the name.
purefa_logging, purefa_ra, purefa_snmp, purefa_alert, purefa_syslog, purefa_ntp, purefa_kmip, purefa_saml, purefa_cbsexpand - read-only listings and state test report changed false and tolerate sparse test results and error responses.
purefa_messages, purefa_audits, purefa_sessions - malformed history, start and end values fail cleanly instead of raising.
purefa_network - state absent of a missing interface is a no-op; interfaces without a gateway are patched with only the changed fields.
purefa_network - newer Purity releases store a virtual interface that is attached to a file server under a name prefixed with the server's name, and refuse a lookup by the bare name. The interface is now found by either the bare name or the prefixed name the array reports, and every update, delete and post-create setting is sent to the name the array stores.
purefa_network - stop asking the array for a subnet named None on every interface creation that does not specify subnet.
purefa_network - the servicelist option was silently ignored on the run that created an interface, and only took effect on a subsequent run. The service list is now sent with the create itself.
purefa_network_access_policy - look rules up by name only (the array rejects names together with policy_names) and verify the owning policy, so a repeated rule delete is a no-op.
purefa_ntp - state test no longer fails on a non-fleet call and state present without a list applies the documented default server.
purefa_pg - state absent on a protection group whose pod no longer exists reports no change instead of failing with Pod ... does not exist.
purefa_pg - a protection group in a pod inside a realm (realm::pod::pg) can be renamed; the module looked up and patched realm::new-name instead of realm::pod::new-name.
purefa_pg - members are added and removed idempotently and targets are only posted when missing.
purefa_pgsched - snap_at, replicate_at and blackout windows are sent in seconds since midnight as the array stores them, and enabled is patched in its own request because the array ignores a frequency or time sent with it; delete uses scaled frequencies.
purefa_pgsnap - add_to_pgs is passed as protection group names (the SDK crashed on references without ids); check mode returns the suffix; plain copies honour with_default_protection.
purefa_pgsnap - state absent on a snapshot whose protection group no longer exists reports no change instead of failing; the failure is kept for every other state and for offloaded snapshots, which can outlive the source group.
purefa_pgsnap - the RETURN documentation describes the suffix key the module returns, not a snapshot key that was never returned.
purefa_phonehome, purefa_maintenance, purefa_snmp_agent, purefa_snmp - repeated runs no longer report a change.
purefa_pod - quota given as a plain number of bytes is applied as documented; it was silently ignored on create and update.
purefa_pod - quota can be cleared, default_protection_pg [] clears the list, state absent on a destroyed pod no longer raises, failover preferences compare by name.
purefa_pod_replica - allow an ActiveCluster stretched pod to be the source of a new replica link on Purity//FA 6.10.5 and later (multi-site continuous replication). Earlier Purity versions get a clear error naming the minimum version, and pausing, resuming or deleting an existing replica link on a stretched pod no longer fails.
purefa_policy - the first NFS rule of a policy is created with every field, nfs, smb and snapshot rules are modified in place, autodir policies rename, user_mapping false is applied, anonuid/anongid use version comparisons, quota_enforced lands in the right field, password_history 0 clears.
purefa_policy_management_access - rename is idempotent.
purefa_realm - QoS can be cleared with 0; a repeated rename is a no-op; error messages name the right option.
purefa_realm - quota given as a plain number of bytes is accepted as documented instead of being converted to 0, which failed realm creation with a misleading minimum-size error and could set a zero quota on update.
purefa_realm - rename_realm() called the nonexistent SDK method patch_realm(), so renaming a realm always failed with AttributeError. It now calls patch_realms().
purefa_saml - credentials are omitted when not given.
purefa_server - network_interfaces compares a requested bare interface name with the server-prefixed name a newer array reports, so a server whose interfaces are already attached no longer reports a change, and an interface stored under another server's name is attached and detached by its stored name.
purefa_smtp - sender, subject_prefix, body_prefix and user can be cleared and state absent clears the relay host with a null instead of the string "None".
purefa_snap - recover_snapshot() called the nonexistent SDK method patch_volume_snapshot(), so recovering a destroyed-but-not-eradicated snapshot always failed with AttributeError. It now calls patch_volume_snapshots().
purefa_snap - a repeated create with the same suffix is a no-op; overwrite copies are checked; offload recovery names are built correctly.
purefa_syslog_settings - a CA certificate can be added when none exists.
purefa_tgroup - the array's error detail is reported; a repeated rename is a no-op.
purefa_timeout - values between 1 and 4 minutes are rejected as documented.
purefa_token - an existing token is detected without creating one; creation reports a change; timeout is honoured on REST 2.
purefa_user - API tokens are only deleted when one exists; a password change no longer skips role, token and key handling.
purefa_user - setting role on a local user whose access comes from a custom management access policy failed with User is not a role user. The user is now moved to the built-in policy of that role.
purefa_vg - a repeated rename no longer re-creates the old group; count without suffix.
purefa_vnc - app names are validated before the deprecation exit; the vnc return value is always a dictionary.
purefa_volume - a repeated rename is a no-op, QoS can be cleared with 0, a bare state present no longer falls into the copy path, add_to_pgs on pod volumes, moving a volume out of a pod.
purefa_volume - accept a volume snapshot as the copy source when target is set, including a protection group snapshot inside a pod such as pod::pgroup.suffix.volume. The module previously treated the snapshot name as a new volume and failed with Size must be specified.
purefa_volume, purefa_vg, purefa_policy - a unit-less size, bw_qos or quota_limit value is treated as bytes instead of 0; a plain-number bw_qos on an existing volume silently removed the bandwidth limit, a plain-number volume size was ignored on update, and vgroup and quota-policy values were sent to the array as 0.
purefa_volume_tags - a kvp value containing a colon (a URL with a port, a timestamp) was truncated at its second colon when tags were created, compared or removed. Only the first colon now separates key from value.
purefa_volume_tags - a copyable-only change of an existing tag is applied.
New Modules
everpure.flasharray.purefa_metrics - Collect OpenMetrics data from a Everpure FlashArray
