Initially we duplicated the code to avoid large scale changes to
the packages. Now we are past metainfo refactor we can remove the
duplication.
Change-Id: I9d0b2756cc6e2a2f4d576afa408a15273a7e1cef
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
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
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
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
* add monkit stat new_remote_segments_needing_repair, which reports the
number of new unhealthy segments in the repair queue since the previous
checker iteration
Change-Id: I2f10266006fdd6406ece50f4759b91382059dcc3
The migration was broken into one migration per table to reduce table locking and reduce the
chances of failure due to SQL timeouts.
Of the 14 fields that lacked time zones, only the 3 named 'interval_start` seemed to have non-UTC data in them.
These fields are fixed in the migration by removing the +00 and adding AT TIME ZONE current_setting('TIMEZONE')
Field with good data are migrated by adding AT TIME ZONE 'UTC'
Note that postgres's timezone() is different than cockroach's timezone() so AT TIME ZONE is used.
https://storjlabs.atlassian.net/browse/SM-104
Change-Id: I410f2f1d7c11b143f17844347f37e6f4b1e70fce
Add a column to the repair queue table in the satellite db for healthy
piece count. When an item is selected from the repair queue, the least
durable segment that has not been attempted in the past hour should be
selected first. This prevents our repairer from getting stuck doing work
on segments that are close to the repair threshold while allowing
segments that are more unhealthy to degrade further.
The migration also clears the repair queue so that the migration runs
quickly and we can properly account for segment health in future repair
work.
We do not select items off the repair queue that have been attempted in
the past six hours. This was changed from on hour to allow us time to
try a wider variety of segments when the repair queue is very large.
Change-Id: Iaf183f1e5fd45cd792a52e3563a3e43a2b9f410b
Currently storage tests were tied to the default lookup limit.
By increasing the limits, the tests will take longer and sometimes
cause a large number of goroutines to be started.
This change adds configurable lookup limit to all storage backends.
Also remove boltdb.NewShared, since it's not used any more.
Change-Id: I1a052f149da471246fac5745da133c3cfc27582e
crdb.ExecuteTx is great, but I don't think it will work right with
PostgreSQL. It works by way of cockroach savepoints, which allows
it to react to retryable errors, whereas tx.Commit() doesn't. But
I don't think PostgreSQL savepoints work exactly the same way. I'm not
100% sure, but it doesn't seem worth the risk.
So, I'm switching one case here to use the new dbutil.WithTx instead,
which will use crdb.ExecuteTx if appropriate. The other case doesn't
need a transaction at all.
Change-Id: I39283f3b5d8d47596db7aff5048bb74597e5918f
Backstory: I needed a better way to pass around information about the
underlying driver and implementation to all the various db-using things
in satellitedb (at least until some new "cockroach driver" support makes
it to DBX). After hitting a few dead ends, I decided I wanted to have a
type that could act like a *dbx.DB but which would also carry
information about the implementation, etc. Then I could pass around that
type to all the things in satellitedb that previously wanted *dbx.DB.
But then I realized that *satellitedb.DB was, essentially, exactly that
already.
One thing that might have kept *satellitedb.DB from being directly
usable was that embedding a *dbx.DB inside it would make a lot of dbx
methods publicly available on a *satellitedb.DB instance that previously
were nicely encapsulated and hidden. But after a quick look, I realized
that _nothing_ outside of satellite/satellitedb even needs to use
satellitedb.DB at all. It didn't even need to be exported, except for
some trivially-replaceable code in migrate_postgres_test.go. And once
I made it unexported, any concerns about exposing new methods on it were
entirely moot.
So I have here changed the exported *satellitedb.DB type into the
unexported *satellitedb.satelliteDB type, and I have changed all the
places here that wanted raw dbx.DB handles to use this new type instead.
Now they can just take a gander at the implementation member on it and
know all they need to know about the underlying database.
This will make it possible for some other pending code here to
differentiate between postgres and cockroach backends.
Change-Id: I27af99f8ae23b50782333da5277b553b34634edc
* pkg/datarepair/repairer: Track always time for repair
Make a minor change in the worker function of the repairer, that when
successful, always track the metric time for repair independently if the
time since checker queue metric can be tracked.
* storage/postgreskv: Wrap error in Get func
Wrap the returned error of the Get function as it is done when the
query doesn't return any row.
* satellite/metainfo: Move debug msg to the right place
NewStore function was writing a debug log message when the DB was
connected, however it was always writing it out despite if an error
happened when getting the connection.
* pkg/datarepair/repairer: Wrap error before logging it
Wrap the error returned by process which is executed by the Run method
of the repairer service to add context to the error log message.
* pkg/datarepair/repairer: Make errors more specific in worker
Make the error messages of the "worker" method of the Service more
specific and the logged message for such errors.
* pkg/storage/repair: Improve error reporting Repair
In order of improving the error reporting by the
pkg/storage/repair.Repair method, several errors of this method and
functions/methods which this one relies one have been updated to be
wrapper into their corresponding classes.
* pkg/storage/segments: Track path param of Repair method
Track in monkit the path parameter passed to the Repair method.
* satellite/satellitedb: Wrap Error returned by Delete
Wrap the error returned by repairQueue.Delete method to enhance the
error with a class and stack and the
pkg/storage/segments.Repairer.Repair method get a more contextualized
error from it.
