Nodes which are offline_suspended will no longer be considered for new
uploads. The current threshold that enters a node into offline
suspension is 0.6. Disqualification for offline suspension is still
disabled.
Change-Id: I0da9abf47167dd5bf6bb21e0bc2186e003e38d1a
Query nodes table using AS OF SYSTEM TIME '-10s' (by default) when on CRDB to alleviate contention on the nodes table and minimize CRDB retries. Queries for standard uploads are already cached, and node lookups for graceful exit uploads has retry logic so it isn't necessary for the nodes returned to be current.
We migrated satelliteDB off of Postgres and over to CockroachDB (crdb), but there was way too high contention for the injuredsegments table so we had to rollback to Postgres for the repair queue. A couple things contributed to this problem:
1) crdb doesn't support `FOR UPDATE SKIP LOCKED`
2) the original crdb Select query was doing 2 full table scans and not using any indexes
3) the SLC Satellite (where we were doing the migration) was running 48 repair worker processes, each of which run up to 5 goroutines which all are trying to select out of the repair queue and this was causing a ton of contention.
The changes in this PR should help to reduce that contention and improve performance on CRDB.
The changes include:
1) Use an update/set query instead of select/update to capitalize on the new `UPDATE` implicit row locking ability in CRDB.
- Details: As of CRDB v20.2.2, there is implicit row locking with update/set queries (contention reduction and performance gains are described in this blog post: https://www.cockroachlabs.com/blog/when-and-why-to-use-select-for-update-in-cockroachdb/).
2) Remove the `ORDER BY` clause since this was causing a full table scan and also prevented the use of the row locking capability.
- While long term it is very important to `ORDER BY segment_health`, the change here is only suppose to be a temporary bandaid to get us migrated over to CRDB quickly. Since segment_health has been set to infinity for some time now (re: https://review.dev.storj.io/c/storj/storj/+/3224), it seems like it might be ok to continue not making use of this for the short term. However, long term this needs to be fixed with a redesign of the repair workers, possible in the trusted delegated repair design (https://review.dev.storj.io/c/storj/storj/+/2602) or something similar to what is recommended here on how to implement a queue on CRDB https://dev.to/ajwerner/quick-and-easy-exactly-once-distributed-work-queues-using-serializable-transactions-jdp, or migrate to rabbit MQ priority queue or something similar..
This PRs improved query uses the index to avoid full scans and also locks the row its going to update and CRDB retries for us if there are any lock errors.
Change-Id: Id29faad2186627872fbeb0f31536c4f55f860f23
We need to be able to list all buckets in DB without knowing project ID.
This method will be used to list buckets for metainfo loop
implementation based on metabase.
Change-Id: Iac75af0eee4f31e80a15577575a8249cbca787b2
We are no longer planning on implementing downtime penalization using
the method described in
docs/blueprints/archive/storage-node-downtime-tracking-deprecated.md.
Now, we are implementing the design described in
docs/blueprints/storage-node-downtime-tracking-with-audits.md.
This change removes the downtime estimation chores from the satellite
core as well as the package satellite/downtime. A future change will
remove the database table.
Change-Id: I1a1d3cf9dceeba36255d25243294865b89925518
Which database access and how it internally does migrations is an
implementation detail and does not belong in the requirements interface.
Change-Id: Ia4a6994f39470063a96a8e5f3a1bd27aa79fe5cd
these are unchanged from storj.io/dbx.
we're importing them because in a later commit we
will change them, and it'd be nice to see that
diff as a separate commit.
Change-Id: I8315130ed6bab397bd65b9a1a90c29d130b8c02d
this change tries really hard to never have all of the storage node
rollups in memory at the same time, up until the rollups are actually
getting summed together.
Change-Id: If67f49e7d71106798d996a6850b3e48671bd9e18
the immediate need is to be able to move the repair queue back out
of cockroach if we can't save it.
Change-Id: If26001a4e6804f6bb8713b4aee7e4fd6254dc326
CRDB doesn't like large deletes. While testing in the POC environment we found that deletes on the serial_numbers table could take hours. This change limits deletes to 1000 at a time (configurable) to avoid blocking other queries.
Change-Id: I08455e25db1574579dd4d7b7125a08e9c913dff1
With the new phase 3 order submission, orders can be added to the
storage and bandwidth rollup tables at timestamps before the most recent
rollup was run. This change shifts the start time of each new rollup
window to account for any unexpired orders that might have been added
since the previous rollup.
A satellitedb migration is necessary to allow upserts in the
accounting_rollups table when entries with identical node_ids and
start_times are inserted.
Change-Id: Ib3022081f4d6be60cfec8430b45867ad3c01da63
We plan to add support for a new Reed-Solomon scheme soon, but our
repair queue orders segments by least number of healthy pieces first.
With a second RS scheme, fewer healthy pieces will not necessarily
correlate to lower health.
This change just adds the new column in a migration. A separate change
will add the new health function.
Right now, since we only support one RS scheme, behavior will not
change. Number of healthy pieces is being inserted as "segment health"
until the new health function is merged.
Segment health is calculated with a new priority function created in
commit 3e5640359. In order to use the function, a new config value is
added, called NodeFailureRate, representing the approximate probability
of any individual node going down in the duration of one checker run.
Change-Id: I51c4202203faf52528d923befbe886dbf86d02f2
It turns out we need to make 2 more changes in order for the new order submission phase 3 to get deployed.
This PR makes 2 changes:
1) when the rollup service deletes tallies, we now keep tallies around until orders expire (vs 1 day like before).
2) the reported rollup chore will now write the storagenode_bandwidth_rollups to a new table _phase2 as an intermediary step so it doesn't conflict with phase 3 order settlement.
These changes need to be deployed for 2 days before we can turn on phase 3 of the new orders settlement workflow.
Change-Id: Iafbff577ba7d55f8f17b7db857311b2ce799de60
With the new overlay.AuditOutcome type for offline audits, the
IsUp field is redundant. If AuditOutcome != AuditOffline, then
the node is online.
In addition to removing the field itself, other changes needed
to be made regarding the relationship between 'uptime' and 'audits'.
Previously, uptime and audit outcome were completely separated. For
example, it was possible to update a node's stats to give it a
successful/failed/unknown audit while simultaneously indicating that
the node was offline by setting IsUp to false. This is no longer possible
under this changeset. Some test which did this have been changed slightly
in order to pass.
Also add new benchmarks for UpdateStats and BatchUpdateStats with different
audit outcomes.
Change-Id: I998892d615850b1f138dc62f9b050f720ea0926b
- Remove flag for switching off offline audit reporting.
- Change the overlay method used from UpdateUptime to BatchUpdateStats, as this
is where the new online scoring is done.
- Add a new overlay.AuditOutcome type: AuditOffline. Since we now use the same
method to record offline audits as success, failure, and unknown, we need to
distinguish offline audits from the rest.
Change-Id: Iadcfe10cf13466fa1a1c2dc542db8994a6423355
This fixes a slow query that was taking up to 4 seconds in production
SELECT node_id, path, piece_num, root_piece_id, durability_ratio, queued_at, requested_at, last_failed_at, last_failed_code, failed_count, finished_at, order_limit_send_count
FROM graceful_exit_transfer_queue
WHERE node_id = '[redacted]'
AND finished_at is NULL
AND last_failed_at is NULL
ORDER BY durability_ratio asc, queued_at asc LIMIT 300 OFFSET 0;
Change-Id: Ib89743ca35f1d8d0a1456b20fa08c683ebdc1549
Use tagsql.DB pointer as step database, to propagate changes
back and forth between actual database and migration.
Adds CreateDB operation to the migration step to be able to
create new dbs before executing migration action.
Adjusts storagenode database migration to use inner tagsql.DB
pointer of each database as step.DB.
Adjusts satellite dabase migration, adds proxy migrationDB field
to satellite db that wraps itself as tagsql.DB, pointer of which
is used as step.DB.
Change-Id: Ifed4de5b01a356cf7b37db64d2eaeb7b61982c5c
This change completes the column migration of
5f6fccc6e8 and
2f648fd981.
It resets every users project limits who are below or equal to our
current production defaults.
Change-Id: Ie041d08bb67b62844f6023190fc00bc2dad5b1cb
Doing it at the ProcessOrders level was insufficient: the endpoints
make multiple database calls. It was a misguided attempt to only
have one spot enter the semaphore. By putting it in the endpoint
we can not only be sure that the concurrency is correctly limited
but it can be configurable easily.
Change-Id: I937149dd077adf9eb87fce52a1a17dc0afe96f64
we have thundering herds of order submissions that take all of the
database connections causing temporary periodic outages. limit
the amount of concurrent order processing to 2.
Change-Id: If3f86cdbd21085a4414c2ff17d9ef6d8839a6c2b
Previously if a node did not have audit history data for each of the
windows over the tracking period, we would give them the benefit of
the doubt and set their score to 1. This was to prevent nodes from
being suspended right out the gate. We need a minimum amount of data
to evaluate them.
However, a node who is actually failing at being online will have no
idea until they have received enough audits and we suspend them.
Instead, we will always use their real score, but use a flag to determine
whether they are eligible for suspension/dq.
Change-Id: I382218f12e8770f95d4bcddcf101ef348940cadf
Repair workers prioritize the most unhealthy segments. This has the consequence that when we
finally begin to reach the end of the queue, a good portion of the remaining segments are
healthy again as their nodes have come back online. This makes it appear that there are more
injured segments than there actually are.
solution:
Any time the checker observes an injured segment it inserts it into the repair queue or
updates it if it already exists. Therefore, we can determine which segments are no longer
injured if they were not inserted or updated by the last checker iteration. To do this we
add a new column to the injured segments table, updated_at, which is set to the current time
when a segment is inserted or updated. At the end of the checker iteration, we can delete any
items where updated_at < checker start.
Change-Id: I76a98487a4a845fab2fbc677638a732a95057a94
