release-5147

What's Changed

• Export db size, deadlocks and changed row metrics by @zaynetro in #7050
• proxy: categorise new cplane error message by @conradludgate in #7057
• pageserver: update swagger for HA APIs by @jcsp in #7070
• fix(metrics): reset TENANT_STATE metric on startup by @koivunej in #7084
• storage controller: refactor non-mutable members up into Service by @jcsp in #7086
• default to tokio-epoll-uring in CI tests & on Linux by @problame in #7077
• fix(tenant/timeline metrics): race condition during shutdown + recreation by @problame in #7064
• fix: Layer delete on drop and eviction can outlive timeline shutdown by @koivunej in #7082
• follow-up(#7077): adjust flaky-test-detection cutoff date for tokio-epoll-uring by @problame in #7090
• fix: warnings when compiling neon extensions by @skyzh in #7053
• Don't return from inside PG_TRY by @save-buffer in #7095
• SIGQUIT instead of SIGKILL prewarmed postgres. by @arssher in #7083
• Keep walproposer alive until shutdown checkpoint is safe on safekepeers by @hlinnaka in #6712
• tests/neon_local: rename "attachment service" -> "storage controller" by @jcsp in #7087
• proxy: cancel http queries on timeout by @conradludgate in #7031
• Forward all backpressure feedback to compute by @petuhovskiy in #7079
• proxy refactor tls listener by @conradludgate in #7056
• storage controller: register nodes in re-attach request by @jcsp in #7040
• Fix max lsn bug by @jbajic in #7007
• pageserver: enable setting stripe size inline with split request. by @jcsp in #7093
• proxy http error classification by @conradludgate in #7098
• proxy: Report warm cold start if connection is from the local cache by @khanova in #7104
• proxy: add new dimension to exclude cplane latency by @khanova in #7011
• fix(metrics): time individual layer flush operations by @koivunej in #7109
• Make tenant_id in TenantLocationConfigRequest optional by @arpad-m in #7055
• throttling: fixup interactions with Timeline::get_vectored by @problame in #7089
• proxy http cancellation safety by @conradludgate in #7117
• pageserver/controller: error handling for shard splitting by @jcsp in #7074
• fix(test suite) virtual_file_io_engine and get_vectored_impl patametrization doesn't work by @problame in #7113
• Bump vendor/postgres v15 v14. by @arssher in #7118
• storage_controller: periodic pageserver heartbeats by @VladLazar in #7092
• test_runner: re-enable large slru benchmark by @VladLazar in #7125
• storage controller: validate DNS of registering nodes by @jcsp in #7101
• doc(README): note cargo-nextest usage by @koivunej in #7122
• spec: disable neon extension auto upgrade by @skyzh in #7128
• proxy: add websocket regression tests by @conradludgate in #7121
• proxy: add more missing warm logging by @khanova in #7133
• docs: sharding phase 1 RFC by @jcsp in #5432
• tests: extend log allow list for test_sharding_split_failures by @jcsp in #7134
• remote_storage: make last_modified and etag mandatory by @jcsp in #7126
• pageserver: remove un-needed "uninit mark" by @jcsp in #5717
• fix(layer): metric miscalculations by @koivunej in #7137
• heavier_once_cell: add detached init support by @koivunej in #7135
• .github: apply timeout to pytest regress by @jcsp in #7142
• docs: shard splitting RFC by @jcsp in #6358
• pageserver: exclude gc_horizon from synthetic size calculation by @jcsp in #6407
• tokio-epoll-uring: fallback to std-fs if not available & not explicitly requested by @problame in #7120
• pageserver: cancellation for remote ops in tenant deletion on shutdown by @jcsp in #6105
• tokio-epoll-uring: use it for on-demand downloads by @problame in #6992
• pageserver, controller: improve secondary download APIs for large shards by @jcsp in #7131
• tokio-epoll-uring: retry on launch failures due to locked memory by @problame in #7141
• fixup(#7120): the macOS code used an outdated constant name, broke the build by @problame in #7150
• Release 2024-03-18 by @vipvap in #7154

New Contributors

• @jbajic made their first contribution in #7007

Full Changelog: release-5090...release-5147