c31348ea9a
Drop the "insert messages" timeout.
2020-10-05 16:57:01 -04:00
c885540749
Check that the return of ZRANGEBYSCORE isn't an empty list.
2020-10-05 10:38:40 -04:00
bb087caddc
Don't panic if a queue exists, but is empty when repairing metadata.
2020-10-04 16:09:56 -04:00
5e3f8b9c2e
Disallow insertion of duplicate messages.
2020-10-04 15:34:14 -04:00
1ccfe928f7
Add a test to make sure that we don't double-insert messages with the same GUID.
2020-10-04 15:34:14 -04:00
3016269268
Revert "Temporarily disable the message persisters entirely."
...
This reverts commit d464721397
2020-10-04 15:25:06 -04:00
952cfae4e6
Repair queue metadata before persisting queues.
2020-10-04 15:25:06 -04:00
df7f209ebc
Revert "Don't insert message batches in transactions."
...
This reverts commit 16eefe333f
2020-10-04 15:12:15 -04:00
d464721397
Temporarily disable the message persisters entirely.
2020-10-04 11:44:35 -04:00
551a85c1e6
Use named variables instead of referring to KEYS/ARGV array indices in message cache scripts.
2020-10-04 11:27:27 -04:00
2686761608
Instrument "get queues to persist" calls and "persist queues" exceptions.
2020-10-04 10:48:42 -04:00
02a2c3224f
Discard unused feature flag constants/mocking.
2020-10-04 10:48:42 -04:00
8ec1dda9ba
Give the persister worker thread a meaningful name.
2020-10-04 10:48:42 -04:00
0308532523
Set a query timeout of 5 seconds when inserting batches of messages.
2020-10-04 10:48:42 -04:00
10b3af2947
Revert "Insert messages individually."
...
This reverts commit 158bfe4816
2020-10-04 10:48:42 -04:00
158bfe4816
Insert messages individually.
2020-10-03 13:13:34 -04:00
16eefe333f
Don't insert message batches in transactions.
2020-10-03 11:43:42 -04:00
65e585e122
Pause only if we're running low on queues to persist.
2020-10-03 11:43:34 -04:00
2ba36ee04c
Add a gauge for worker thread liveness.
2020-10-03 11:43:34 -04:00
fc05529574
Let MessagePersister manage its own worker thread.
2020-10-03 11:43:34 -04:00
07d24f487a
Don't re-register metrics for shared circuit breakers.
2020-10-02 15:05:00 -04:00
811acdb7f5
Use separate namespaces for Redis breaker/retry metrics.
2020-10-02 10:57:05 -04:00
a7266364d1
Refactor peer pruning to be more retry-friendly.
2020-10-01 17:17:07 -04:00
e83b41dc01
Reduce default Redis cluster command timeout to 3 seconds.
2020-10-01 17:17:07 -04:00
76665dd56e
Retry Redis commands that time out.
2020-10-01 17:17:07 -04:00
2d42b478ba
Consolidate cluster and pub/sub circuit breakers.
2020-10-01 17:17:07 -04:00
885fa6beae
Add tests for Device#isEnabled.
2020-10-01 12:54:35 -04:00
65cdd5fcbe
Drop the 365-day check when deciding if an account is enabled.
2020-10-01 12:54:35 -04:00
4302e19aba
Register a UUID argument factory for the messages database.
2020-10-01 11:06:43 -04:00
0c6f05f34a
Add a (failing!) test for sending a sealed-sender message after a non-sealed-sender message.
2020-10-01 11:06:43 -04:00
8040c285cd
Include stack traces when reporting persistence issues.
2020-09-30 11:47:16 -04:00
ada454f56f
Add a meter for persisting individual messages.
2020-09-30 10:39:56 -04:00
57d2ef8740
Return queues to the "to persist" list if something goes wrong during persistence.
2020-09-30 10:39:56 -04:00
a97e0982e3
Add an integration test for message persistence.
2020-09-30 10:39:56 -04:00
eaa2060d84
Fix an incorrect locking key and some previously-suppressed lock contention issues.
2020-09-30 10:39:56 -04:00
3e02c574e7
Log exceptions when persisting messages.
2020-09-30 10:39:56 -04:00
c7230ccbb0
Remove messages from the cache in bulk.
2020-09-29 10:58:02 -04:00
fc71ced660
Persist messages in batches.
2020-09-29 10:58:02 -04:00
6041a9d094
Make exit conditions slightly more conservative.
2020-09-29 10:58:02 -04:00
599cd766e1
Let Dropwizard manage persister thread lifecycles.
2020-09-29 10:58:02 -04:00
e64c8007c0
Detect GV2 capability in non-gcm Android devices
2020-09-28 15:54:10 -04:00
9339823e84
Add temporary metrics to monitor the ratio of enabled/disabled accounts.
2020-09-28 15:33:52 -04:00
e6d4620af1
Only allow linking desktop clients if they support the third-generation GV2 capability.
2020-09-25 17:08:32 -04:00
656e6db846
Only consider desktop devices GV2-capable if they send the third-gen GV2 capability.
2020-09-25 17:08:32 -04:00
30474e3a2b
Add a test for message ordering.
2020-09-25 11:41:58 -04:00
460bd98f1b
Add metrics for messages missing GUIDs.
2020-09-25 11:41:22 -04:00
a553eba574
Add an API endpoint for deleting accounts.
2020-09-25 11:39:17 -04:00
61f515670c
Add plumbing for deleting accounts and all associated data.
2020-09-25 11:39:17 -04:00
789af0f8a6
Add support for deleting keys associated with an account.
2020-09-25 11:39:17 -04:00
86fae58c96
Add support for deleting account entities from the database.
2020-09-25 11:39:17 -04:00
c54d3abe47
Check for the second-gen GV2 capability when linking devices.
2020-09-24 19:04:02 -04:00
6fe511eb50
Fix a bad size check when loading stored messages.
2020-09-23 18:02:33 -04:00
17d18b22c7
Drop pub/sub sending logic from WebsocketSender.
2020-09-23 14:51:02 -04:00
66a04ed730
Don't explicitly notify clients when messages get persisted.
2020-09-23 14:51:02 -04:00
7e14a0bc30
Drop pub/sub operations from WebsocketConnection.
2020-09-23 14:51:02 -04:00
77de0f86dc
Require desktop clients to send the new gv2-2 capability flag.
2020-09-23 12:05:58 -04:00
3b4bc9163a
Untangle thread pool names, tweak sizes, and add instrumentation.
2020-09-22 10:21:33 -04:00
e146135bd1
Don't attempt to send more messages if sending failed for any reason.
2020-09-22 10:21:33 -04:00
e9e18afb4a
Add a (failing) integration test demonstrating an infinite loop.
2020-09-22 10:21:33 -04:00
62c31eb202
Revert "Revert keyspace delivery for all messages"
...
This reverts commit 4dc49604b6
2020-09-22 10:21:33 -04:00
1eacee85ae
Count how many iOS users set the old GV2 capability flag.
2020-09-21 18:58:07 -04:00
5986145282
Add a second-generation GV2 capability and ignore the old capability for iOS devices.
2020-09-21 18:57:53 -04:00
b134a69a28
Record the number of authentications for users with/without GV2 support.
2020-09-21 15:42:13 -04:00
83f9eacac4
Refactor UserAgentTagUtil to parse UA strings with UserAgentUtil.
2020-09-21 12:24:08 -04:00
baab6b951b
Add a general utility class for parsing user-agent strings.
2020-09-21 12:24:08 -04:00
903a1bec91
Reject (eventually) oversize messages.
2020-09-17 17:07:20 -04:00
ebc3a251b7
Drop the UUID addressing capability flag entirely.
2020-09-14 15:36:29 -04:00
a567f4a6de
Don't check UUID capability when blocking capability downgrades.
2020-09-14 15:36:29 -04:00
4dc49604b6
Revert keyspace delivery for all messages
...
* Revert "Send all messages via keyspace notifications when a feature flag is enabled."
This reverts commit fadcf62166c02b25576689788fa665a052e2ee8f158e5004b76f9ff3be37f766c577438f53152c3e7bbc88d71668256d2343
2020-09-14 15:35:10 -04:00
fadcf62166
Send all messages via keyspace notifications when a feature flag is enabled.
2020-09-11 13:12:17 -04:00
c02b255766
Consolidate semaphore release logic.
2020-09-11 13:12:17 -04:00
89788fa665
Represent stored message state as an enumeration rather than a collection of booleans.
2020-09-11 13:12:17 -04:00
a052e2ee8f
Refactor: collapse state into semaphores/atomic booleans.
2020-09-11 13:12:17 -04:00
158e5004b7
Refactor: move sendNextMessagePage into its own method.
2020-09-11 13:12:17 -04:00
6f9ff3be37
Avoid querying the database if we think all new messages are in the cache.
2020-09-11 13:12:17 -04:00
f766c57743
Query for more stored messages if an update happens while we're already processing a batch.
2020-09-11 13:12:17 -04:00
8f53152c3e
Only send the "queue cleared" message once per websocket session.
2020-09-11 13:12:17 -04:00
7bbc88d716
Let processStoredMessages handle requery logic.
2020-09-11 13:12:17 -04:00
68256d2343
Only allow one thread to process stored messages at a time.
2020-09-11 13:12:17 -04:00
f88c440c48
Automatically retry when Twilio returns unreachable ( #190 )
...
* Parse and log the Twilio error code
* Automatically retry without sender ID when Twilio returns unreachable
* Remove attempt count and pass around whether or not sender id was used
2020-09-10 13:58:39 -05:00
cfa56ba6d4
Remove the "send online messages via keyspace notifications" feature flag.
2020-09-10 10:41:20 -04:00
2c6b646d87
Enforce no capability downgrade on device verification
2020-09-09 16:05:00 -04:00
e7572094b5
Require all enabled devices to support GV2.
2020-09-09 16:05:00 -04:00
5e34823a49
Optionally send online-only messages via keyspace notifications.
2020-09-09 14:42:09 -04:00
fdef21a871
Record and listen for ephemeral messages in a separate queue.
2020-09-09 14:42:09 -04:00
d40cff8a99
Revert "Add a system for storing, retrieving, and notifying listeners about ephemeral (online) messages."
...
This reverts commit 06754d6158
2020-09-08 15:55:09 -04:00
8927e45ded
Revert "Optionally send online-only messages via keyspace notifications."
...
This reverts commit 12fe28d8ab
2020-09-08 15:55:09 -04:00
1a93df92d4
Replace DeliveryStatus with a simple boolean.
2020-09-08 11:29:33 -04:00
12fe28d8ab
Optionally send online-only messages via keyspace notifications.
2020-09-08 11:19:55 -04:00
06754d6158
Add a system for storing, retrieving, and notifying listeners about ephemeral (online) messages.
2020-09-08 11:14:42 -04:00
1d5087374e
Jettison UUID-or-E164 plumbing in favor of UUID-only.
2020-09-08 09:30:47 -04:00
8356264fe0
Rename RedisClusterMessagesCache and related classes to just MessagesCache.
2020-09-08 09:30:47 -04:00
18ecd748dd
Entirely discard the old message cache machinery.
2020-09-08 09:30:47 -04:00
e324f27655
Stop sending/processing CONNECTED pub/sub messages.
2020-09-03 13:52:43 -04:00
afd645fb11
Retrieve messages using commands available in Redis 3.
2020-09-03 13:31:55 -04:00
5b42593fbb
Persist messages one page at a time.
2020-09-03 12:08:46 -04:00
5c04f2634a
Use a dedicated executor service for dispatching keyspace notifications.
2020-09-03 11:04:48 -04:00
ad01610d1e
Rely on the client presence manager to decide whether to send push notifications.
2020-09-03 11:04:48 -04:00
697c380cd1
Close websocket connections when displaced.
2020-09-03 11:04:48 -04:00
81e8143a43
Rely solely on the clustered message cache.
2020-09-02 11:57:33 -04:00
8409986ef5
Mirror persistence operations from the new persister to the old persister.
2020-09-02 11:02:40 -04:00
2b50367d7f
Put message persisters behind feature flags.
2020-09-02 11:02:40 -04:00
1dcc491fec
Move cache-mirroring operations to the calling thread.
2020-09-01 12:34:37 -04:00
d715f86713
Refactor to constants
2020-09-01 10:55:26 -04:00
5221828705
Increase maximum sticker size to 300 kibibytes
...
In preparation for animated stickers, allow stickers to be up to 300
kibibytes.
2020-09-01 10:55:26 -04:00
6aa4acd3db
Mirror "clear queue" operations to the clustered cache.
2020-09-01 10:55:07 -04:00
15936c29c1
Let Dropwizard manage the lifecycle of the feature flag manager.
2020-09-01 10:50:59 -04:00
8b70c69a0d
Replace metrics with logging statements.
2020-08-31 15:57:17 -04:00
dfe80a30dc
Make ScourMessageCacheCommand a ConfiguredCommand instead of an EnvironmentCommand.
2020-08-31 15:57:17 -04:00
ce026e7ad0
Don't send contacts to CDS if they've opted out of discoverability. (SERVER-130)
2020-08-27 15:58:02 -04:00
58e3122dab
Add a discoverableByPhoneNumber account attribute. (SERVER-129)
2020-08-27 15:58:02 -04:00
3b55b2d1b2
Actually make the "scour message cache" available to Dropwizard. Oops.
2020-08-27 15:15:04 -04:00
2326e61de5
Clear and re-create gauges to avoid "stuck" feature flag reporting.
2020-08-27 13:18:12 -04:00
32b18c9509
Add an endpoint for getting the current state of feature flags.
2020-08-27 13:18:12 -04:00
acf52ad8a3
Make feature flag manager tests use a real database to avoid over-mocking.
2020-08-27 13:18:12 -04:00
08dd493f98
Don't report exceptions as part of traffic metrics.
2020-08-27 13:17:57 -04:00
07bbe7dfb2
Return to an async model for push notification latency.
2020-08-27 10:51:44 -04:00
0aa1b80e3e
Add a command for persisting any detached messages in the old message cache.
2020-08-27 10:51:12 -04:00
5ac390281e
Add an abstract base class for Redis singleton tests.
2020-08-27 10:51:12 -04:00
ac465c5a18
Add a Lettuce-based Redis singleton client.
2020-08-27 10:51:12 -04:00
1ef3546822
Add support for server-side feature flags
2020-08-26 20:27:33 -04:00
e74ad2b555
Make RedisClusterMessagesCache a Managed class.
2020-08-25 10:58:01 -04:00
71c0056c66
Use lots of specific subscriptions instead of one monster subscription to minimize load.
2020-08-25 10:58:01 -04:00
56b27ea785
Record experiment outcomes with timers instead of counters.
2020-08-25 10:57:44 -04:00
2d75f59d33
Add support for UUID-only delivery certificates. (SERVER-132)
2020-08-20 17:05:53 -04:00
a709a3bcc0
Remove a candidate metric provider.
2020-08-20 15:40:56 -04:00
34bf5112e0
Drop TimeProvider.
2020-08-20 15:40:24 -04:00
bfe18d1d28
Re-nerf the clustered message persister.
2020-08-20 15:38:09 -04:00
6a76afc20d
Add a test to make sure the persister is respecting persist delays.
2020-08-20 15:38:09 -04:00
9c469c2f96
Base persister tests on a real Redis cluster.
2020-08-20 15:38:09 -04:00
2ab42f3dd6
Refine and expand clustered message cache metrics.
2020-08-19 11:39:05 -04:00
af34b43a8d
Reactivate the message notification experiment.
2020-08-19 11:39:05 -04:00
0f71cc7864
Rename metrics associated with cluster circuit breakers for clarity.
2020-08-18 17:59:00 -04:00
df90de3a5f
Change default Lettuce command timeout to 10s.
2020-08-18 16:21:42 -04:00
42ea7a9814
Revert Lettuce connection pooling.
2020-08-18 16:21:42 -04:00
c683cbdb2d
Time Redis operations.
2020-08-18 12:20:12 -04:00
d243b73678
Make Lettuce connection pools configurable. Double the default size.
2020-08-18 12:20:12 -04:00
dc28d063aa
Reactivate the explicit client presence experiment.
2020-08-17 11:34:27 -04:00
bb6045c1d0
Disarm the client presence manager experiment.
2020-08-15 20:23:05 -04:00
f1a74b5939
Disarm new message keyspace notifications.
2020-08-15 20:23:05 -04:00
6fb9038af1
Move to a synchronous, pooled connection model for Redis clusters.
2020-08-14 17:15:56 -04:00
27f721a1f5
Update to resilience4j 1.5.0.
2020-08-14 17:15:56 -04:00
5717dc294e
Combine the read/write breakers for Redis clusters.
2020-08-14 17:15:56 -04:00
ae0f8df11b
Break out FaultTolerantPubSubConnection as its own thing so different use cases can have their own subscription space.
2020-08-14 17:15:56 -04:00
77460ba502
Remove keyspace notification configuration checks because AWS doesn't support `CONFIG GET`.
2020-08-13 15:32:25 -04:00
f8235da4d8
Fix an issue where the queue for a thread pool was not bounded.
2020-08-13 12:46:11 -04:00
8d3316ccd6
Listen for new messages via keyspace notifications.
2020-08-13 12:17:04 -04:00
2c29f831e8
Add an explicit client presence system.
2020-08-13 10:56:26 -04:00
9457325119
Add pub/sub affordances to FaultTolerantRedisCluster.
2020-08-13 10:56:26 -04:00
189f8afcc9
Warm up the test cluster before running tests to avoid transient startup jitters.
2020-08-13 10:56:26 -04:00
9699b67510
Record the size of outgoing message lists.
2020-08-12 16:57:10 -04:00
d60633a46c
Add a meter for the number of messages we send via websocket connections.
2020-08-12 16:57:10 -04:00
0fcf28e7e7
Use the MessagesManager to actually persist messages.
2020-08-11 15:50:22 -04:00
5fad8f74b1
Factor MessagePersister into its own class.
2020-08-11 15:50:22 -04:00
e35e34d2e0
Move operation-mirroring logic to MessagesManager.
2020-08-11 15:50:22 -04:00
31a215d4d6
Use "global." instead of "g." as the prefix for global config options.
2020-08-11 11:55:35 -04:00
30948de13d
Update a metric provider dependency and remove a workaround for an upstream issue.
2020-08-11 11:02:38 -04:00
b97158bf7b
Create global remote config controllable in the signal server configuration ( #127 )
...
* Add global config controller through file rather than database
* Do no permit attempting to set or delete global config entries
2020-08-10 16:31:15 -05:00
6646be8d94
Make CpuUsageGauge a CachedGauge.
2020-08-10 12:56:37 -04:00
647a2aea64
Cache a reference to the OS management bean to avoid repeated lookups.
2020-08-10 12:56:37 -04:00
58e58ce51c
Remove a candidate metric provider.
2020-08-10 11:03:20 -04:00
4b7e48d3ec
Override default ingestion URI for SignalFx ( #131 )
2020-08-07 15:29:42 -05:00
0e074d3a5a
Copy SignalFxMeterRegistry into a new class to get better logging
2020-08-07 16:01:56 -04:00
ea00224e7f
Add support for reporting metrics to signalfx ( #129 )
2020-08-07 11:10:31 -05:00
38293efe75
Keep a running count of the number of open websockets.
2020-08-06 16:07:34 -04:00
3286c5e174
Disable Redis persistence for tests.
2020-08-06 11:22:51 -04:00
e0f8a28f38
Close connections before closing the whole cluster client.
2020-08-06 11:22:31 -04:00
bf1b00b163
Drop a spurious RedisClusterClient.
2020-08-06 11:22:31 -04:00
4fa3a136ad
Remove arbitrary SMS and add a NANPA message service ( #123 )
...
* Remove arbitrary SMS code
This code has run its course and is no longer needed for now.
* Add elements to sample config that were left out
* Add a messaging service for NANPA
* Fixup sample config capitalization
2020-08-05 13:35:11 -05:00
178a6bd66e
Log the top-level exception name and message when crawling badness happens.
2020-08-05 11:23:16 -04:00
57e1339230
Further restrict user agent pattern matching ( #120 )
...
* Further restrict user agent pattern matching
* Add static qualifier to method
2020-08-04 12:58:16 -05:00
4144423227
Publish percentiles for Micrometer distributions/timers.
2020-08-04 10:58:59 -04:00
4d03514142
Add a command for clearing the messages cache cluster.
2020-08-04 10:58:41 -04:00
0bc5566976
Mirror delete-after-persist operations to the clustered message cache.
2020-08-04 10:58:41 -04:00
925567add5
Actually "plug in" the reglock counter.
2020-08-03 15:43:33 -04:00
ad97731d46
Reduce the maximum number of versions in play to 1,000.
2020-08-03 15:42:15 -04:00
40684a93a2
Restrict user-agent version matching to a more confined space.
2020-08-03 15:42:15 -04:00
f3b644ceb8
Update the push latency manager to use UUIDs and a Redis cluster.
2020-08-03 15:36:02 -04:00
901ba6e87f
Added a push latency manager.
2020-08-03 15:36:02 -04:00
76389bd584
Clear would-be-persisted messages from the cache cluster, but don't store them to the database.
2020-07-30 19:14:39 -04:00
7bf8650d59
Un-manage FaultTolerantRedisCluster so it shuts down at JVM shutdown instead of Jetty shutdown.
2020-07-30 18:37:38 -04:00
7cb24dd96d
Add environment tag to datadog metric reporting
2020-07-30 18:04:16 -04:00
dee040318a
Add the host tag to datadog metric reporting
2020-07-30 18:04:16 -04:00
baf563e46d
Temporarily disarm the actual persisting part of the message persister.
2020-07-30 17:12:37 -04:00
e10246f10b
Use Dropwizard timers/histograms for persister metrics.
2020-07-30 14:27:06 -04:00
a9dfd88671
Start the clustered message persister at application startup.
2020-07-30 12:32:35 -04:00
beac73b6c8
Add a cluster-capable message persister
2020-07-30 11:39:14 -04:00
f9f93c77e2
Use UUIDs instead of phone numbers as account identifiers in clustered message cache
2020-07-30 11:39:14 -04:00
6fc1b4c6c0
Add a cluster-backed message cache.
2020-07-30 11:39:14 -04:00
639898ec07
Expand Experiment to deal with async suppliers and Optionals.
2020-07-30 11:39:14 -04:00
3d3790fdbc
Add binary execution methods to ClusterLuaScript.
2020-07-30 11:39:14 -04:00
69c8968cb0
Add byte-array-based methods to FaultTolerantRedisCluster.
2020-07-30 11:39:14 -04:00
aa25fc7901
Fix UsernamesManager metric/logger names.
2020-07-29 11:00:29 -04:00
4aba493ee2
Fix the key used for database crawler workers.
2020-07-29 10:58:06 -04:00
b9cfac5934
Introduce additional metric aggregators.
2020-07-28 15:11:51 -04:00
f8e97fcc32
revise 12 hour active user fudge to 8 hours for better continuity of data from a month ago
2020-07-28 11:09:41 -07:00
7f8f2641f6
Simplify registration lock counting by avoiding inactive accounts.
2020-07-28 11:48:20 -04:00
022dbb606f
Count registration lock versions when crawling the account database.
2020-07-28 11:48:20 -04:00
fea72b190d
Record message content size as a dimensioned distribution.
2020-07-28 11:47:56 -04:00
eea073f882
Decommission the old cache.
2020-07-28 10:29:28 -04:00
Jon Chambers