Show CAS mount renewal progress in system.cas_mounts

[filimonov]

While a CAS mount lease renewal retries, system.cas_mounts showed nothing: an aging last_renewed_at, lifecycle still Live, and the attempt counters moved only after the call ended. An operator could not tell a renewal retrying through an object-store outage from a renewal thread that had died. Three columns now report the renewal as it runs.

Why it matters for the lease

The next change of the series (#2431) makes the renewal retry until the store gives a definitive answer, so a live holder will spend tens of seconds — or longer — inside one renewal call. Without these columns that pause has no visible sign of life; with them, SELECT lifecycle, renewal_in_progress, renewal_attempts, last_renewal_error FROM system.cas_mounts during an outage reads Live / 1 / 7 / "put gc/server-roots/…/mount: Timeout …". The same fields were the missing evidence in the incident analysis of #2431.

Changes

• system.cas_mounts: renewal_in_progress (a renewal call is running), renewal_attempts (physical PUTs sent by the current call), last_renewal_error (verb, key and error class of the last refused request; no credentials, no bodies). Appended at the end; existing queries are unaffected. Documented in docs/en/operations/system-tables/cas_mounts.md.
• Request engine: a RequestObserver with on_write_attempt and on_failure callbacks, called per sent attempt and per refused request.
• Renewal thread: a RenewalProgress record under a leaf mutex (renewal_progress_mutex, never held across I/O or allocation beyond one string move); in_progress is reset by SCOPE_EXIT on every exit, exceptions included. CASMountRenewalAttempts/CASMountRenewalRetries are counted per attempt instead of from the result.
• A local failure of the resolve read (not a request to the store) is not published.

Risks / notes

• Nothing changes on the store or in the protocol. Cost: one leaf-mutex lock per attempt, at one renewal per cas_mount_renew_period_ms.
• The system-table read copies the record under the lock and releases it before its own LIST of the slots.

Testing

• Unit: the engine observer receives attempts and failures; renewalProgress follows the call and resets on an exception; the local-failure filter (red on the base: gtest_cas_request_observer.cpp, gtest_cas_mount_renewal_progress.cpp).
• Integration: test_cas_mount_renewal_retry gains a scenario that reads the columns while RustFS is stopped.
• CAS* gate: 2614 (release), 2618 (ASan, no error).

Related: #2431

Changelog category (leave one):

• Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

system.cas_mounts gains renewal_in_progress, renewal_attempts and last_renewal_error, updated while a CAS mount lease renewal runs.

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)

Updated docs/en/operations/system-tables/cas_mounts.md and docs/en/antalya/cas/architecture/mounts-and-leases.md.

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Unit tests
• Performance tests
• Aarch64 tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All Regression
• Disable CI Cache

Regression jobs to run:

• Fast suites (mostly <1h)
• Aggregate Functions (2h)
• Alter (1.5h)
• Benchmark (30m)
• CAS (content-addressed storage; Antalya only)
• ClickHouse Keeper (1h)
• Iceberg (2h)
• LDAP (1h)
• OAuth (5m)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [ba17785]

[filimonov]

Stateless tests (amd_asan_ubsan, cas s3 storage, parallel, 1/2): Server died with Logical error: GroupingAggregatedTransform pushed bucket A twice. The stack is in aggregation (MergingAggregatedMemoryEfficientTransform), no CAS frames; the same failure hits other PRs on this base. Not related to this PR. Tracked upstream: ClickHouse#115986