Thread resolving is one of the few tasks that isn't retried on failure.
One common cause for failure of this task is a well-connected user replying to
a toot from a little-connected user on a small instance: the small instance
will get many requests at once, and will often fail to answer requests within
the 10 seconds timeout used by Mastodon.
This changes makes the ThreadResolveWorker retry a few times, with a
rapidly-increasing time before retries and large random contribution in order
to spread the load over time.
* Clean up reblog-tracking sets from FeedManager
Builds on #5419, with a few minor optimizations and cleanup of sets
after they are no longer needed.
* Update tests, fix multiply-reblogged case
Previously, we would have lost the fact that a given status was
reblogged if the displayed reblog of it was removed, now we don't.
Also added tests to make sure FeedManager#trim cleans up our reblog
tracking keys, fixed up FeedCleanupScheduler to use the right loop,
and fixed the test for it.
* Keep references to all reblogs of a status on home feed
When inserting reblog: Add to set of reblogs of this status on
the feed, if original status was present in the feed, add it to
that set as well.
When removing a reblog: Remove it from that set. Take random
remaining item from the set. If one exists, re-insert it into feed,
otherwise do not re-insert anything.
Fix#4210
* When original is removed, toss out reblog references
- Rename Mastodon::TimestampIds into Mastodon::Snowflake for clarity
- Skip for statuses coming from inbox, aka delivered in real-time
- Skip for statuses that claim to be from the future
* Improve error handling on LinkCrawlWorker
* Ignore TimeoutError and InvalidURIError too
* Record errors to debug log
* Enable dead job queue on LinkCrawlWorker
Since most of acceptable errors were already ignored, only our side issue should go to dead job queue.
* Ignore all http gem errors
* Revert "Enable UniqueRetryJobMiddleware even when called from sidekiq worker (#4836)"
This reverts commit 6859d4c028.
* Revert "Do not execute the job with the same arguments as the retry job (#4814)"
This reverts commit be7ffa2d75.
* Whenever a remote keypair changes, unfollow them and re-subscribe to them
In Mastodon (it could be different for other OStatus or AP-enabled software),
a keypair change is indicative of whole user (or instance) data loss. In this
situation, the “new” user might be different, and almost certainly has an empty
followers list. In this case, Mastodon instances will disagree on follower
lists, leading to unreliable delivery and “shadow followers”, that is users
believed by a remote instance to be followers, without the affected user
knowing.
Drawbacks of this change are:
1. If an user legitimately changes public key for some reason without losing
data (not possible in Mastodon at the moment), they will have their remote
followers unsubscribed/re-subscribed needlessly.
2. Depending of the number of remote followers, this may generate quite some
traffic.
3. If the user change is an attempt at usurpation, the remote followers will
unknowingly follow the usurper. Note that this is *not* a change of
behavior, Mastodon already behaves like that, although delivery might be
unreliable, and the usurper would not have known the former user's
followers.
* Rename ResubscribeWorker to RefollowWorker
* Process followers in batches
When a new user confirms their e-mail, bootstrap their home timeline
by automatically following a set of accounts. By default, all local
admin accounts (that are unlocked). Can be customized by new admin
setting (comma-separated usernames, local and unlocked only)
* Decouple Status#local? from uri being nil
* Replace on-the-fly URI generation with stored URIs
- Generate URI in after_save hook for local statuses
- Use static value in TagManager when available, fallback to tag format
- Make TagManager use ActivityPub::TagManager to understand new format
- Adjust tests
* Use other heuristic for locality of old statuses, do not perform long query
* Exclude tombstone stream entries from Atom feed
* Prevent nil statuses from landing in Pubsubhubbub::DistributionWorker
* Fix URI not being saved (#4818)
* Add more specs for Status
* Save generated uri immediately
and also fix method order to minimize diff.
* Fix alternate HTML URL in Atom
* Fix tests
* Remove not-null constraint from statuses migration to speed it up
Requires moving Atom rendering from DistributionWorker (where
`stream_entry.status` is already nil) to inline (where
`stream_entry.status.destroyed?` is true) and distributing that.
Unfortunately, such XML renderings can no longer be easily chained
together into one payload of n items.
* Add handling of Linked Data Signatures in payloads
* Add a way to sign JSON, fix canonicalization of signature options
* Fix signatureValue encoding, send out signed JSON when distributing
* Add missing security context
* ActivityPub migration procedure
Once one account is detected as going from OStatus to ActivityPub,
invalidate WebFinger cache for other accounts from the same domain
* Unsubscribe from PuSH updates once we receive an ActivityPub payload
* Re-subscribe to PuSH unless already unsubscribed, regardless of protocol
*Note: OStatus URIs are invalid for ActivityPub. But we have them for
as long as we want to keep old OStatus-sourced content and as long as
we remain OStatus-compatible.*
- In Announce handling, if object URI is not a URL, fallback to object URL
- Do not use specialized ThreadResolveWorker, rely on generalized handling
- When serializing notes, if parent's URI is not a URL, use parent's URL
* Add ActivityPub inbox
* Handle ActivityPub deletes
* Handle ActivityPub creates
* Handle ActivityPub announces
* Stubs for handling all activities that need to be handled
* Add ActivityPub actor resolving
* Handle conversation URI passing in ActivityPub
* Handle content language in ActivityPub
* Send accept header when fetching actor, handle JSON parse errors
* Test for ActivityPub::FetchRemoteAccountService
* Handle public key and icon/image when embedded/as array/as resolvable URI
* Implement ActivityPub::FetchRemoteStatusService
* Add stubs for more interactions
* Undo activities implemented
* Handle out of order activities
* Hook up ActivityPub to ResolveRemoteAccountService, handle
Update Account activities
* Add fragment IDs to all transient activity serializers
* Add tests and fixes
* Add stubs for missing tests
* Add more tests
* Add more tests
* Do not raise unretryable exceptions in ResolveRemoteAccountService
* Removed fatal exceptions from ResolveRemoteAccountService
Exceptions that cannot be retried should not be raised. New exception
class for those that can be retried (Mastodon::UnexpectedResponseError)
* Wrap methods of ProcessFeedService::ProcessEntry in classes
This is a change same with 425acecfdb, except
that it has the following changes:
* Revert irrelevant change in find_or_create_conversation
* Fix error handling for RemoteActivity
* Introduce Ostatus name space
* Add dependency on idn-ruby to speed up URI normalization
* Use normalized_host instead of normalize.host when applicable
When we are only interested in the normalized host, calling normalized_host
avoids normalizing the other components of the URI as well as creating a
new object
* Add Request class with HTTP signature generator
Spec: https://tools.ietf.org/html/draft-cavage-http-signatures-06
* Add HTTP signature verification concern
* Add test for SignatureVerification concern
* Add basic test for Request class
* Make PuSH subscribe/unsubscribe requests use new Request class
Accidentally fix lease_seconds not being set and sent properly, and
change the new minimum subscription duration to 1 day
* Make all PuSH workers use new Request class
* Make Salmon sender use new Request class
* Make FetchLinkService use new Request class
* Make FetchAtomService use the new Request class
* Make Remotable use the new Request class
* Make ResolveRemoteAccountService use the new Request class
* Add more tests
* Allow +-30 seconds window for signed request to remain valid
* Disable time window validation for signed requests, restore 7 days
as PuSH subscription duration (which was previous default due to a bug)
Since there is little point in retrying so often when a service is down
or does not exist anymore. Subscriptions are renewed 1 day before they
should expire, so retrying in 30 minutes, then 2 hours, then 12 hours
is fine. If even after that, the remote server does not work, there is
little sense in retrying more often than once a day
Also, uniqueness of the job should ensure that failed retries will
not result in multiple retries for the same endpoint when the next
resubscription cycle comes
* Add overview of active sessions
* Better display of browser/platform name
* Improve how browser information is stored and displayed for sessions overview
* Fix test
* Fix#2347 - Bind web UI access token to session
When you logout, session also destroys the access token, so it's no longer
valid. If access token is destroyed some other way, the session is also
destroyed, requiring a re-login.
Fix#1681 - Add scheduler to remove revoked access tokens and grants
* Fix test
* Introduce domains method to Account relation
Account had followers_domains method, which was excessively specific.
Let relation of Account have domains method instead.
* Move follow_mapping in Account to AccountInteractions
* Introduce shared examples for AccountAvatar inclusion
* Cover Account more
* Make Pubsubhubbub::DistributionWorker handle both single stream entry
arguments, as well as arrays of stream entries
* Add BatchedRemoveStatusService, make SuspendAccountService use it
* Improve method names
* Add test
* Add more tests
* Use PuSH payloads of 100 to have a clear mapping of
1000 input statuses -> 10 PuSH payloads
It was nice while it lasted
* Add form for account deletion
* If avatar or header are gone from source, remove them
* Add option to have SuspendAccountService remove user record, add tests
* Exclude suspended accounts from search
* Do not cancel PuSH subscriptions after encountering "permanent" error response
After talking with MMN about it, turns out some servers/php setups do
return 4xx errors while rebooting, so this anti-feature that was meant
to take load off of the hub is doing more harm than good in terms of
breaking subscriptions
* Update delivery_worker.rb
* Fix#2473 - Use sidekiq scheduler to refresh PuSH subscriptions instead of cron
Fix an issue where / in domain would raise exception in TagManager#normalize_domain
PuSH subscriptions refresh done in a round-robin way to avoid hammering a single
server's hub in sequence. Correct handling of failures/retries through Sidekiq (see
also #2613). Optimize Account#with_followers scope. Also, since subscriptions
are now delegated to Sidekiq jobs, an uncaught exception will not stop the entire
refreshing operation halfway through
Fix#2702 - Correct user agent header on outgoing http requests
* Add test for SubscribeService
* Extract #expiring_accounts into method
* Make mastodon:push:refresh no-op
* Queues are now defined in sidekiq.yml
* Queues are now in sidekiq.yml
* Fix#2119 - Whenever about to send a HTTP request, normalize the URI
* Add test for IDN request in FetchLinkCardService
* Perform IDN normalization on domains before they are stored in the DB
* Make private toots get PuSHed to subscription URLs that belong to domains where you have approved followers
* Authorized followers controller, stub for bulk action
* Soft block in the background
* Add simple test for new controller
* Rename Settings::FollowersController to Settings::FollowerDomainsController, paginate results,
rename "private" post setting to "followers-only", fix pagination style, improve post privacy
preferences style, improve warning style
* Extract compose form warnings into own container, show warning when posting to followers-only with unlocked account
* Fix#1141, fix#1126 - Work through UpdateRemoteProfileService for both <feed> and <entry> top-level tags
* Improve code quality, remove line unrelated to fix
* Rewrite Atom generation from stream entries to use Ox instead of Nokogiri::Builder
StreamEntry is now limited to only statuses, which allows some optimization. Removed
extra queries on AccountsController#show. AtomSerializer instead of AtomBuilderHelper
used in AccountsController#show, StreamEntriesController#show, StreamEntryRenderer
and PubSubHubbub::DistributionWorker
PubSubHubbub::DistributionWorker moves n+1 DomainBlock query to PubSubHubbub::DeliveryWorker
instead.
All Salmon slaps that aren't based on StreamEntry still use AtomBuilderHelper and Nokogiri
* All Salmon slaps now use Ox instead of Nokogiri. No touch from status on account
must be added to the Sidekiq invokation in your systemd file
The pull queue will handle link crawling, thread resolving, and OStatus
processing. Such tasks are more likely to hang for a longer time (due to
network requests) so it is more sensible to not make the "in-house" tasks
wait for them.
This is a big one, so let me enumerate:
Accounts as well as stream entry pages now contain Link headers that
reference the Atom feed and Webfinger URL for the former and Atom entry
for the latter. So you only need to HEAD those resources to get that
information, no need to download and parse HTML <link>s.
ProcessFeedService will now queue ThreadResolveWorker for each remote
status that it cannot find otherwise. Furthermore, entries are now
processed in reverse order (from bottom to top) in case a newer entry
references a chronologically previous one.
ThreadResolveWorker uses FetchRemoteStatusService to obtain a status
and attach the child status it was queued for to it.
FetchRemoteStatusService looks up the URL, first with a HEAD, tests
if it's an Atom feed, in which case it processes it directly. Next
for Link headers to the Atom feed, in which case that is fetched
and processed. Lastly if it's HTML, it is checked for <link>s to the Atom
feed, and if such is found, that is fetched and processed. The account for
the status is derived from author/name attribute in the XML and the hostname
in the URL (domain). FollowRemoteAccountService and ProcessFeedService
are used.
This means that potentially threads are resolved recursively until a dead-end
is encountered, however it is performed asynchronously over background jobs,
so it should be ok.
ThibG