Feature/team events #35
Conversation
| rsp <- Data.deleteTeam tid | ||
| when ((view teamBinding . Data.tdTeam <$> team) == Just Binding) $ | ||
| journal $ Journal.teamDelete tid | ||
| return rsp |
There was a problem hiding this comment.
Shouldn't we reverse this, i.e., deleteTeamshould probably be done after the journaling as that'd guarantee that we journal the event at least once
There was a problem hiding this comment.
Makes sense in this case.
|
|
||
| teamCreate :: TeamId -> UserId -> IO TeamEvent | ||
| teamCreate tid uid = do | ||
| now <- utcTimeToInt64 <$> liftIO getCurrentTime |
There was a problem hiding this comment.
Instead of requiring a new library, how about using utcTimeToPOSIXSeconds which is part of time?
> n <- getCurrentTime
> n
2017-07-26 12:52:28.373924 UTC
> utcTimeToInt64 n
1501073548
> round (utcTimeToPOSIXSeconds n) :: Int64
1501073548
| rsp <- addTeamMemberInternal tid Nothing Nothing nmem mems | ||
| let membs = nmem^.ntmNewTeamMember : mems | ||
| let bUsers = view userId <$> filter (`hasPermission` SetBilling) membs | ||
| journal $ Journal.teamUpdate tid (fromIntegral $ length membs) bUsers |
There was a problem hiding this comment.
I noticed that
let membs = filter (\u -> u^.userId /= remove) mems
let bUsers = view userId <$> filter (`hasPermission` SetBilling) membs
$ Journal.teamUpdate tid (fromIntegral $ length membs) bUsers
is a recurring pattern, with only slight variation on how membs is constructed. Can we refactor that?
| mEnv <- view aEnv | ||
| for_ mEnv $ \e -> do | ||
| event <- liftIO ev | ||
| void $ Aws.execute e (Aws.enqueue event) |
There was a problem hiding this comment.
I think you don't need this void here
| either (throwM . GeneralError) (const (return ())) res | ||
| where | ||
| event = decodeLatin1 $ B64.encode $ encodeMessage e | ||
| req url = SQS.sendMessage url event & SQS.smMessageGroupId .~ Just "team.events" |
There was a problem hiding this comment.
Just nitpicking here but should we define this somewhere else since it should be used by consumers too?
There was a problem hiding this comment.
Are you referring to the MessageGroupId? This is only relevant when sending to a FIFO queue, consumers don't need to know about it:
Receiving Messages
You can't request to receive messages with a specific message group ID.
jschaul
force-pushed
the
feature/team-events
branch
13 times, most recently
from
3d8d59e to
827858c
Compare
Introduces journaling via AWS SQS for team events (create, update, delete). * Team deletions become soft deletes (to allow for eventual consistency on the journal consumer side: a cassandra table scan allows for re-creating create/delete events) * types-common-journal depends on `protoc` at compile time for protobuf code generation. * Journaling in galley is optional at run time (via optional `---team-events-queue-name` and `--aws-region` params) and at integration test time (via an optional `GALLEY_SQS_TEAM_EVENTS` environment variable) to allow galley to function without SQS. * integration tests consume SQS events, making them slower (from ~90s to ~110s) (to be improved)
jschaul
force-pushed
the
feature/team-events
branch
from
827858c to
621b5fb
Compare
* ListResponse is 1-indexed, not 0-indexed * implement patch in terms of get and put * Implements patch _only_ for userName, displayName, externalId * Add Azure acceptance test Co-authored-by: fisx <mf@zerobuzz.net>
…s components (#35) * add a first scenario explaining cert renewal for a 3-master K8s v1.14 setup installed with Kubespray * add additional structure to reflect (fingers crossed) future k8s content * typo * Minor changes, add one paragraph Co-authored-by: jschaul <jschaul@users.noreply.github.com>
Introduces journaling via AWS SQS for team events (create, update, delete).
protocat compile time for protobuf code generation.---team-events-queue-nameand--aws-regionparams) and at integration test time (via an optionalGALLEY_SQS_TEAM_EVENTSenvironment variable) to allow galley to function without SQS.