8506ca7919
Merge pull request #12998 from ieQu1/dev/improve-latency
...
Use leader's clock when calculating LTS cutoff timestamp
2024-05-13 21:54:06 +02:00
3da3a36863
test(ds): Add generation in the replication suite
2024-05-13 19:51:04 +02:00
9f7ef9f34f
fix(ds): Apply review remarks
2024-05-13 19:35:24 +02:00
07aa708894
test(ds): Refactor replication suite
2024-05-09 03:56:56 +02:00
63e51fca66
test(ds): Use streams to fill the storage
2024-05-09 02:46:57 +02:00
a0a3977043
feat(ds): Assign latest timestamp deterministically
2024-05-08 23:17:57 +02:00
2236af84ba
feat(ds): two-stage storage commit on the storage level
2024-05-08 23:17:57 +02:00
1ddbbca90e
feat(ds): Allow incremental update of the LTS trie
2024-05-08 23:17:57 +02:00
68ca891f41
test(ds): Use streams to create traffic
2024-05-08 23:17:57 +02:00
d84c180ccb
feat(dsrepl): avoid contacting unreachable ra servers
...
Assuming estabilished Erlang distribution channel is a reliable way to
tell whether a remote node is reachable.
2024-05-08 18:12:13 +02:00
3642bcd1b6
docs(ds): Fix comment for the builtin DS metrics
2024-05-06 11:21:32 +02:00
b2a633aca1
fix(ds): Use leader's clock for computing LTS safe cutoff time
2024-05-06 11:21:32 +02:00
1ff2e02fd9
feat(ds): Pass current time to the storage layer via argument
2024-05-06 11:21:32 +02:00
8ac9700aab
feat(ds): Add an API for DB-global variables
2024-05-06 11:21:32 +02:00
86d45522e3
fix(dsrepl): Don't reverse elements of batches
2024-05-06 11:21:32 +02:00
bcfa7b2209
fix(ds): Destroy LTS tries when the generation is dropped
2024-05-06 11:21:32 +02:00
9999ccd36c
feat(ds): Ignore safe cutoff time for streams without varying levels
2024-05-06 11:21:32 +02:00
e4c3283c9c
docs(ds): Update README with CLI and REST API endpoints
2024-04-23 16:28:35 +02:00
4c76a2574d
fix(ds): Fix egress flush condition
2024-04-21 21:51:31 +02:00
43f8346c00
fix(dssnap): ensure idempotent write of empty chunks
2024-04-19 18:52:33 +02:00
93bb840365
docs(ds): Update README
2024-04-17 01:21:52 +02:00
5d7b2e2ce6
fix(dsrepl): attempt leadership transfer on terminate
...
In addition to on removal. The reasoning is basically the same: try to
avoid situations when log entries are replicated (or will be considered
replicated when the new leader is elected) but the leader terminates
before replying to the client.
To be clear: this is a stupid solution. Something much more robust is
needed.
2024-04-15 22:05:24 +02:00
89f42f1171
fix(dsrepl): make placeholder shard process permanent under supervisor
2024-04-15 16:43:52 +02:00
c4d1360b96
fix(dsrepl): trigger election for new ra servers unconditionallly
...
Otherwise we might end up in a situation when there's no member online
yet at the time of the election trigger, and the election will never
happen.
2024-04-15 16:42:29 +02:00
d12e907209
fix(dsrepl): correctly handle ra membership change command results
...
Before this change, results similar to `{error, {no_more_servers_to_try,
[{error, nodedown}, {error, not_member}]}}` were considered retryable
failures, which is incorrect.
2024-04-08 22:44:34 +02:00
3223797ae5
fix(dsrepl): attempt leadership transfer before server removal
...
This should make it much less likely to hit weird edge cases that lead
to duplicate Raft log entries because of client retries upon receiving
`shutdown` from the leader being removed.
2024-04-08 22:43:58 +02:00
1e95bd4da6
test(dsrepl): test unresponsive nodes removal / node restarts
2024-04-08 21:27:56 +02:00
7a836317ac
fix(dsrepl): trigger unfinished shard transition upon startup
...
Also provide a trivial API to trigger them by hand.
2024-04-08 16:12:42 +02:00
75bb7f5cdc
fix(dsrepl): retry only `{add, Site}` crashed membership transitions
...
To minimize the potential negative impact of removal transitions that
crash for some unknown and unusual reasons.
2024-04-08 16:04:33 +02:00
4c0cc079c2
fix(dsrepl): apply unnecessary rebalancing transitions cleanly
2024-04-08 13:25:45 +02:00
dcde30c38a
test(dsrepl): add two more testcases for rebalancing
2024-04-08 13:22:31 +02:00
2ace9bb893
chore(dsrepl): sprinkle few comments and typespecs for exports
2024-04-07 22:51:56 +02:00
ecaad348a7
chore(dsrepl): update few outdated comments / TODOs
2024-04-07 22:51:56 +02:00
6293efb995
fix(dsrepl): retry crashed membership transitions
2024-04-07 22:51:56 +02:00
826ce5806d
fix(dsrepl): ensure that new member UID matches server's UID
...
Before that change, UIDs supplied in the `ra:add_member/3` were not
the same as those servers were using. This haven't caused any issues
for some reason, but it's better to ensure that UIDs are the same.
2024-04-07 22:31:24 +02:00
556ffc78c9
feat(dsrepl): implement membership changes and rebalancing
2024-04-05 18:57:28 +02:00
d6058b7f51
feat(dsrepl): allow to subscribe to DB metadata changes
...
Currently, only shard metadata changes are announced to the
subscribers.
2024-04-05 17:40:55 +02:00
a07295d3bc
fix(ds): address shards in the supervisor properly
2024-04-05 17:40:38 +02:00
a62db08676
feat(ds): Add REST API for durable storage
2024-04-05 15:22:06 +02:00
d09787d1a6
fix(ds): Fix return types in replication_layer_meta
2024-04-05 15:22:06 +02:00
70396e9766
Merge pull request #12825 from keynslug/feat/EMQX-12110/repl-meta-api
...
feat(dsrepl): add APIs to manage DB replication sites
2024-04-04 22:32:03 +02:00
df6c5b35fe
feat(dsrepl): add more primitive operations to modify DB sites
2024-04-04 21:22:49 +02:00
bb8ffee18c
feat(dsrepl): add API to get current DB replication sites
2024-04-04 21:22:02 +02:00
ad52f7838e
feat(dsrepl): add APIs to manage DB replication sites
2024-04-04 21:22:01 +02:00
c57c36adb2
feat(ds): clear all checkpoints when (re)starting storage layer
...
Fixes https://emqx.atlassian.net/browse/EMQX-12143
2024-04-04 14:05:52 -03:00
f37ed3a40a
fix(ds): Limit the number of retries in egress to 0
2024-04-03 16:38:49 +02:00
2bbfada7af
fix(ds): Make async batches truly async
2024-04-03 11:57:47 +02:00
92ca90c0ca
fix(ds): Improve egress logging
2024-04-03 11:57:47 +02:00
ae5935e7f7
test(ds): Attempt to stabilize metrics_worker tests in CI
2024-04-02 19:14:10 +02:00
4382971443
fix(ds): Preserve errors in the egress
2024-04-02 16:47:43 +02:00
94ca7ad0f8
feat(ds): Report counters for LTS storage layout
2024-04-02 16:47:43 +02:00
b379f331de
fix(sessds): Handle errors when storing messages
2024-04-02 16:47:41 +02:00
f41e538526
feat(sessds): Observe next time
2024-04-02 16:45:52 +02:00
75b092bf0e
fix(ds): Actually retry sending batch
2024-04-02 16:45:49 +02:00
0de255cac8
feat(ds): Report egress flush time
2024-04-02 16:25:04 +02:00
044f3d4ef5
fix(ds): Don't reverse entries in the atomic batch
2024-04-02 16:25:04 +02:00
606f2a88cd
feat(ds): Add egress metrics
2024-04-02 16:25:04 +02:00
c9de336234
feat(ds): Add metrics worker to the builtin db supervision tree
2024-04-02 16:25:04 +02:00
778e897f1f
chore(dsrepl): describe snapshot ownership and few shortcomings
2024-04-02 13:48:51 +02:00
c666c65c6a
test(ds): factor out storage iteration into helper module
2024-04-02 13:48:51 +02:00
7cebf598a8
chore(dsrepl): simplify snapshot transfer code a bit
...
Co-Authored-By: Thales Macedo Garitezi <thalesmg@gmail.com>
2024-04-02 13:48:51 +02:00
e029b8f996
test(dsrepl): wait for whole cluster readiness
...
To minimize the chance of flaky tests due to the shards not being
completely online.
Co-Authored-By: Thales Macedo Garitezi <thalesmg@gmail.com>
2024-04-02 13:48:50 +02:00
e8b06a6a9f
chore(dsrepl): mark few more BPAPI targets as obsolete
2024-04-02 13:48:50 +02:00
d31cd0c728
feat(ds): ensure LTS state ids are deterministic
2024-04-02 13:48:50 +02:00
2cd357a5bd
fix(ds): ensure store batch is idempotent wrt generations
2024-04-02 13:48:50 +02:00
77a022bd93
feat(dsrepl): transfer storage snapshot during ra snapshot recovery
2024-04-02 13:48:49 +02:00
b8b9b7739b
chore(ds): slightly simplify working with storage generations
2024-04-02 13:48:08 +02:00
fa66a640c3
fix(dsrepl): handle RPC errors gracefully when storage is down
2024-03-28 15:17:01 +01:00
db9efb9317
chore: bump apps versions
2024-03-28 10:19:09 +01:00
796c04e7a8
test: fix flaky test
...
We should emit the trace event before replying to callers.
Example failure:
https://github.com/emqx/emqx/actions/runs/8378977952/job/22946318696#step:6:182
```
=CRITICAL REPORT==== 21-Mar-2024::17:45:37.676024 ===
"check stage" failed: error
{assertMatch,[{module,emqx_ds_storage_bitfield_lts_SUITE},
{line,270},
{expression,"? of_kind ( emqx_ds_replication_layer_egress_flush , Trace )"},
{pattern,"[ # { batch := [ _ , _ , _ ] } ]"},
{value,[]}]}
Stacktrace: [{emqx_ds_storage_bitfield_lts_SUITE,
'-t_atomic_store_batch/1-fun-1-',1,
[{file,
"/__w/emqx/emqx/apps/emqx_durable_storage/test/emqx_ds_storage_bitfield_lts_SUITE.erl"},
{line,270}]},
{emqx_ds_storage_bitfield_lts_SUITE,t_atomic_store_batch,1,
[{file,
"/__w/emqx/emqx/apps/emqx_durable_storage/test/emqx_ds_storage_bitfield_lts_SUITE.erl"},
{line,249}]}]
```
2024-03-21 15:47:29 -03:00
68af211130
fix(ds): reply sync callers after raft store failure
2024-03-21 15:40:21 -03:00
70737a437a
fix(ds): add caller to pending replies before flushing
2024-03-21 14:39:21 -03:00
fe50a1711b
fix(ds-egress): drop pending batch on failures
...
Before this commit, messages in the current batch will be retried as
part of next batch. This could have led to message duplication which is
probably not what the user wants by default.
2024-03-20 13:20:25 +01:00
a1f5de3f5b
fix(dsrepl): turn memoize into a safer function
2024-03-20 13:20:24 +01:00
d39ca41070
chore(dsrepl): mark per-node `add_generation` RPC target obsolete
...
Also annotate internal exports with comments according with their
intended use.
2024-03-20 13:20:24 +01:00
35b18f9125
fix(dsrepl): properly handle error conditions in generation mgmt
...
Also update few outdated typespecs. Also make error reasons easier
to comprehend.
2024-03-20 13:20:24 +01:00
f2268aa69a
fix(dsrepl): use correct base dir for ra system stuff
...
Co-Authored-By: Thales Macedo Garitezi <thalesmg@gmail.com>
2024-03-20 13:20:24 +01:00
404e919494
refactor(dsrepl): make shard allocator more robust and consistent
...
Co-Authored-By: Thales Macedo Garitezi <thalesmg@gmail.com>
2024-03-20 13:20:24 +01:00
0e18bd6e80
fix(dsrepl): increase replication site id bitsize back
...
In order to minimize chances of site id collision to practically zero.
2024-03-20 13:20:24 +01:00
ac9700dd28
fix(dsrepl): split shard allocator into a separate module
2024-03-20 13:20:23 +01:00
1b647035d0
chore(dsrepl): make dialyzer a bit happier
2024-03-20 13:20:23 +01:00
611b3f0e07
feat(dsrepl): use more straightforward way to drop ra shards
2024-03-20 13:20:23 +01:00
74881e8706
feat(dsrepl): make storage layer unaware of granularity of time
...
Storage also becomes a bit more pure, depending on the upper layer to
provide the timestamps, which also makes it possible to handle more
operations idempotently.
2024-03-20 13:20:23 +01:00
3cb36a5619
feat(ds-lts): extract timestamp from storage key itself
...
1. This avoids the need to deserialize the message to get the timestamp.
2. It also makes possible to decouple the storage key timestamp from the
message timestamp, which might be useful for replication purposes.
2024-03-19 20:21:56 +01:00
5cc0246351
feat(dsrepl): allow to tune select ra options
2024-03-19 20:21:55 +01:00
54b5adf868
feat(dsrepl): allocate shards predictably
...
To ensure strictly optimal and fair shard allocation across
cluster. Before this commit it was quite easy to end up with
an allocation significantly skewed towards some node, because
of the nature of randomness and relatively small number of
shards.
2024-03-19 20:21:55 +01:00
887e151be5
fix(dsrepl): handle errors gracefully in shard egress process
...
Also add cooldown on timeout / unavailability.
2024-03-19 20:21:53 +01:00
e16aee99b4
chore(dsrepl): clarify how to perform leadership transfer in runtime
2024-03-19 20:21:18 +01:00
00d509f27b
feat(dsrepl): prefer local replica in read path
...
To optimize out any unnecessary RPCs. Given the load should be
smoothed evenly across the cluster, choosing non-leader node is
not a priority.
2024-03-19 20:11:42 +01:00
19305c223c
fix(dsrepl): require majority for replication-related tables
2024-03-19 20:11:42 +01:00
f89909f60c
fix(dsrepl): tolerate trigger election timeouts for existing servers
2024-03-19 20:11:42 +01:00
3b59cf2ebf
feat(dsrepl): move shard allocation to separate process
...
That starts shard and egress processes only when shards are fully
allocated.
2024-03-19 20:11:41 +01:00
4dafbf21f6
fix(dsrepl): make db + shard part of machine state
...
It doesn't feel right, but right now is the easiest way to have it
in the scope of `apply/3`, because `init/1` doesn't actually invoked
for ra machines recovered from the existing log / snapshot.
2024-03-19 20:11:41 +01:00
d19128ed65
feat(dsrepl): cache shard metadata in persistent terms
2024-03-19 20:11:41 +01:00
e6c2c2fb07
feat(dsrepl): manage generations / db config through ra machine
2024-03-19 20:11:39 +01:00
5e94bdb932
feat(dsrepl): allocate shards once predefined number of sites online
...
Before this commit the most likely shard allocation outcome was
that all shard are allocated to just one node.
2024-03-19 20:11:03 +01:00
be793e4735
fix(dsrepl): reassign timestamp at the time of submission
...
This is needed to ensure total message order for a shard, and
guarantee that no messages will be written "in the past". which
may break replay consistency.
2024-03-19 20:11:01 +01:00
146f082fdc
feat(dsrepl): implement raft-based replication
...
Still very rough but mostly working.
2024-03-19 20:09:44 +01:00
f2dc940436
Merge remote-tracking branch 'upstream/release-56' into 0319-sync-release56
2024-03-19 15:20:08 +01:00
11ae04d810
test(ds): rm unused var warning
2024-03-14 17:16:24 -03:00
2ebc8dcc55
fix(ds): use `infinity` timeout when storing batches
2024-03-14 10:17:18 -03:00
6af01b916e
feat(ds): implement `get_delete_streams`, `make_delete_iterator` and `delete_next` callbacks for builtin storage
...
Part of https://emqx.atlassian.net/browse/EMQX-11841
2024-03-08 09:56:46 -03:00
f7e3afde16
test(ds): avoid introducing new macros
2024-03-07 16:49:20 +01:00
69427dc42d
test(ds): add tests for error mapping and replay recovery
2024-03-07 12:59:58 +01:00
09905d78cd
chore(ds): make error handling slightly simpler
...
Co-Authored-By: Thales Macedo Garitezi <thalesmg@gmail.com>
2024-03-07 12:59:57 +01:00
b39c710ec2
fix(ds): tidy up few typespecs
2024-03-07 12:59:57 +01:00
2146d9e1fe
feat(ds): introduce error classes in critical API functions
...
For now, only recoverable / unrecoverable errors are introduced.
2024-03-07 12:59:57 +01:00
5d87d400f4
feat(ds): add atomic store API
...
Part of https://emqx.atlassian.net/browse/EMQX-11841
2024-03-06 15:24:14 -03:00
06334798a5
fix(ds): fix `drop_generation` typespec
...
This typespec fix will be used downstream by other backends.
2024-03-04 14:15:59 -03:00
b706caf294
feat(ds): export types
2024-02-29 14:27:18 +03:00
d5ae0e5c53
feat(ds): update delete/count interface
2024-02-28 22:51:24 +03:00
b010d34640
chore(ds): add delete callbacks
2024-02-26 17:35:13 +03:00
46877e979b
chore: update copyright-year
2024-02-23 08:21:06 +01:00
d469f4158e
chore: bump app vsns
2024-02-20 16:53:57 -03:00
8cfb22f0b8
fix(ds): Retry getting the shard leader
2024-02-16 12:42:48 +01:00
280fcd8c52
Merge pull request #12437 from ieQu1/dev/optimize_make_filter
...
Optimize emqx_ds_bitmask_keymapper:make_filter function.
2024-02-05 17:32:28 +01:00
4665837cf0
fix(ds): Apply review remarks
2024-02-05 16:52:06 +01:00
c7888ad1f1
Merge pull request #12475 from ieQu1/dev/lean-stream
...
Use a more compact data structure to represent streams
2024-02-05 13:55:24 +01:00
698ba3f271
fix(ds): Optimize emqx_ds_bitmask_keymapper:make_filter
...
This optimization makes idle polling faster
2024-02-05 10:54:19 +01:00
8edbec5929
refactor(ds): Clarify the language used in ds_bitmapper
2024-02-05 10:54:18 +01:00
75023f2ca3
Merge pull request #12442 from ieQu1/dev/ds-license-apache
...
Relicense apps/emqx_durable_storage under Apache 2.0
2024-02-05 10:16:52 +01:00
2e56810ea2
refactor(ds): Use a simple improper list to represent the streams
2024-02-03 21:15:54 +01:00
3d2ac97c61
feat(ds): Add a CLI interface to inspect status of DS databases
2024-02-02 10:11:01 +01:00
139f5dc3bd
chore(ds): Remove an obsolete document; superceded by ./README.md
2024-02-01 00:14:08 +01:00
b50d6bf1fd
chore(ds): Change the license to Apache 2.0
...
Due to technicalities parts of the original code were licensed under
BSL.
In preparations for the public release of the feature, the license has
been changed to Apache 2.0
2024-02-01 00:10:48 +01:00
f7b12470dd
test(ds): fix inter-suite flakiness
...
Attempt to mitigate this frequent source of flakiness:
```
=CRASH REPORT==== 31-Jan-2024::17:30:15.025404 ===
crasher:
initial call: emqx_ds_replication_layer_egress:init/1
pid: <0.11312.0>
registered_name: []
exception error: no match of right hand side value {error,
no_leader_for_shard}
in function emqx_ds_replication_layer_egress:init/1 (/emqx/apps/emqx_durable_storage/src/emqx_ds_replication_layer_egress.erl, line 93)
in call from gen_server:init_it/2 (gen_server.erl, line 980)
in call from gen_server:init_it/6 (gen_server.erl, line 935)
ancestors: [<0.11310.0>,<0.11304.0>,emqx_ds_builtin_databases_sup,
emqx_ds_builtin_sup,emqx_ds_sup,<0.11236.0>]
message_queue_len: 0
messages: []
links: [<0.11310.0>]
dictionary: []
trap_exit: true
status: running
heap_size: 376
stack_size: 28
reductions: 231
neighbours:
```
2024-01-31 15:42:26 -03:00
d51deac222
fix(ds): use configured data dir for site storage
2024-01-31 15:42:26 -03:00
35c4ef2ee2
Merge pull request #12407 from zmstone/0126-script-to-update-bsl-license-convert-time
...
0126 script to update bsl license convert time
2024-01-30 09:54:34 +08:00
7ee0a6aaa4
Merge pull request #12414 from ieQu1/dev/ds-readme
...
doc(ds): Update README
2024-01-29 18:01:03 +01:00
82403167c2
chore: update BSL license change date
2024-01-29 16:47:31 +01:00
2479e1189a
fix(ds): Remove unused module
2024-01-29 00:36:13 +01:00
96c527541b
docs(ds): Update README
2024-01-28 20:27:20 +01:00
eec56b0d6b
fix(sessds): Improve comments
2024-01-26 17:49:33 +01:00
8e8d3af096
fix(sessds): Refactor emqx_persistent_session_ds to use CRUD module
2024-01-26 17:49:33 +01:00
8e31afe6c2
fix(ds): don't make data dir part of the schema
...
The data directory was ending up being persisted in the database schema. This led to
issues when opening the DB on different nodes.
2024-01-25 14:44:06 -03:00
305a54f646
chore(ds): Update BPAPI version
2024-01-24 19:33:30 +01:00
eee221f1d0
feat(ds): Make egress batching configurable
2024-01-24 19:33:30 +01:00
137535a821
feat(ds): Introduce egress process for the builtin backend
2024-01-24 19:33:30 +01:00
9b7df302e8
fix(ds): Cache database metadata in RAM
2024-01-24 18:43:48 +01:00
1eb47d0c16
perf(ds): inherit only LTS paths containing wildcards when adding a new generation
...
Fixes https://github.com/emqx/emqx/pull/12338#discussion_r1462139499
2024-01-23 09:20:28 -03:00
d122340c13
Merge pull request #12338 from thalesmg/ds-message-gc-20240115
...
feat(ds): add message GC
2024-01-22 16:57:26 -03:00
db710c4be5
feat(lts): inherit previous generation's lts when possible
2024-01-22 14:53:17 -03:00
75b08b525b
feat(ds): add `list_generations` and `drop_generation` APIs
2024-01-22 14:53:17 -03:00
57074015c6
feat(ds): allow customizing the data directory
...
The storage expectations for the RocksDB DB may be different from our usual data
directory. Also, it may consume a lot more storage than other data.
This allows customizing the data directory for the builtin DS storage backend.
Note: if the cluster was already initialized using a directory path, changing that config
will have no effect. This path is currently persisted in mnesia and used when reopening
the DB.
2024-01-19 13:07:24 -03:00
7763b0fd34
chore: bump app vsns
2024-01-08 17:42:56 -03:00
756980837c
Merge branch 'release-54' into sync-r54-m-20240108
2024-01-08 17:39:42 -03:00
caf461fdf6
fix(ds): Don't start the supervision tree when feature is not in use
2024-01-08 17:59:19 +01:00
9e0f3ce53b
feat(ds): restore original add_generation/update_db_config callback semantics
2023-12-29 13:12:15 +03:00
5e100f52b8
style: erlfmt all `rebar.config` files and `bin/nodetool`
2023-12-29 09:08:03 +08:00
23ded313ec
chore: update app versions
2023-12-22 15:29:22 +01:00
ieQu1