rate
Graceful exit is very slow at the moment. Over the last couple days we
increase the batch size on Stefans satellite to 1000 but as a side
effect the error rate was increased. With a batch size of 500 the error
rate looks stable.
This PR will increase the default to batch size to 300. Graceful exit
will still be painful slow but at least it will be a bit faster. At the
same time this PR also increases the number of errors we tolerate. We
don't want to DQ slow storage nodes just because they didn't finish all
300 transfers in time. We want to give them more retries.
Change-Id: I92e3f99e116d4988457d8b902a88e85ed1bcc1a7
Fixes https://storjlabs.atlassian.net/browse/USER-240
- Adds UnsynchronizedPut method to metainfo service that overwrites any
existing pointer under the same path
- Uses UnsynchronizedPut in the metainfo endpoint for committing the
segments
Change-Id: Icb43f31ea33f14066ca9dfdcf226eb3079b90948
if redis crashed in the middle of tally we could have a situation
where we erroneously subtract from a project total. Currently,
`latest` should never be less than `initial`
Change-Id: Ibb5ab724ac0ad4d684f7954fad7a9e061104b7df
Currently SNs report their free disk space once per hour. If a node
becomes full, it has to wait until the next contact cycle begins to
report; all the while receiving and failing upload requests. By increasing
the minimum required disk space, we can give the storage nodes more time
to report their space before the completely fill up. This change goes
hand-in-hand with another change we want to implement: trigger capacity
report on SN immediately upon falling below threshold.
Change-Id: I12f778286c6c3f582438b0e2949765ac43325e27
This change resolves all the storage node addresses to their IP addresses
before giving them to the uplink so that the uplink doesn't have to resolve
a hundred hosts and can immediately connect to improve uplink performance.
Change-Id: Idb834351e0fece409d74c8a1c29b0b8c9b09c9ff
This peer will contain our administrative panels.
It's completely separated from our other satellite
processes because it allows better control for restricting
access to it.
Change-Id: Ifca473bee82ff6c680b346918ba32b835a7a6847
Add monkit metric for the rate-limit when the rate limit is hit
Logs warning with projectID
https://storjlabs.atlassian.net/browse/SM-165
Change-Id: I352dc40006021990d1bc66a999f62bbf8deb54db
Adds ExcludedIPs to the NodeCriteria for selecting new storage
nodes. Previously, ExcludedIPs was only added to the NodeCriteria
for selecting reputable storage nodes. Now that both are included
in the FindStorageNodesWithPreferences call, it should no longer
be possible to repair pieces to nodes that are on the same IP as
nodes already storing pieces from that segment.
Adds TestSelectNewStorageNodesExcludedIPs to make sure that
SelectNewStorageNodes returns nodes with different IP addresses.
https://storjlabs.atlassian.net/browse/V3-3011
Change-Id: Ic2d5e607cadeba6e8d5c40f9717149cb30880335
more trustworthy downtime tracking
Detection chore: Do not update downtime at all from the detection chore.
We only want to include downtime between two explicitly failed ping attempts
(the duration between last contact success and the first failed ping is no longer
included in downtime calculation)
Estimation chore: If the satellite started after the last failed ping for a node,
do not include offline time since the last failed ping time - only
estimate based on two failed pings with no satellite downtime in
between.
This protects us from including satellite downtime in our storagenode downtime calculations.
Change-Id: I1fddc9f7255a7023e02474255d70c64faae75b8a
Sometimes the upload that is supposed to fail due to excess usage
would pass. This looks to be because it's overwriting another object
uploaded earlier in the test and deleting the old pointer. If tally
happened to run after the pointer is deleted but before the current
upload reaches the live accounting check, it might pass through.
The solution is to upload to a different path each time.
Change-Id: Ie6c825b9c6eab9ed53426ae262e7997bcb6beb7f
In the methods we use to retrieve a user's chargeable BW, we were summing GET, GET_AUDIT,
and GET_REPAIR. We only want to charge for GET
Change-Id: Icead7695494b22c7c835482cf8b1512a980d59f1
this commit updates our monkit dependency to the v3 version where
it outputs in an influx style. this makes discovery much easier
as many tools are built to look at it this way.
graphite and rothko will suffer some due to no longer being a tree
based on dots. hopefully time will exist to update rothko to
index based on the new metric format.
it adds an influx output for the statreceiver so that we can
write to influxdb v1 or v2 directly.
Change-Id: Iae9f9494a6d29cfbd1f932a5e71a891b490415ff
it was noticed that if you had a long lived transaction A that
was blocking some other transaction B and A was being aborted
due to retriable errors, then transaction B was never given
priority. this was due to using savepoints to do lightweight
retries.
this behavior was problematic becaue we had some queries blocked
for over 16 hours, so this commit addresses the issue with two
prongs:
1. bound the amount of time we will retry a transaction
2. create new transactions when a retry is needed
the first ensures that we never wait for 16 hours, and the value
chosen is 10 minutes. that should be long enough for an ample
amount of retries for small queries, and huge queries probably
shouldn't be retried, even if possible: it's more preferrable to
find a way to make them smaller.
the second ensures that even in the case of retries, queries that
are blocked on the aborted transaction gain priority to run.
between those two changes, the maximum stall time due to retries
should be bounded to around 10 minutes.
Change-Id: Icf898501ef505a89738820a3fae2580988f9f5f4
Allow rate limit project cache to expire so we can make project level rate limit changes without restarting the satellite process.
Change-Id: I159ea22edff5de7cbfcd13bfe70898dcef770e42
Core shouldn't be handling any repair load and we have already disabled it in production.
Let's make it official and remove it.
Change-Id: I46e236692a9164421648cfc974dd3246416b2e00
Satellite now is keeping RS values for uplink but old uplinks were using
default bucket settings. Because of that we need to override buckets
settings with satellite settings to avoid breaking older uplinks.
Change-Id: Ia1068db70e4adbf741c5e81d27d9e39799049c22
before dbx would generate a compilcated blob of conditions that
encoded a row comparison, which only optimized to an index seek
on cockroachdb. this means that sqlite and postgres both had
quadratic behavior on paged queries of this form. instead, use
the implicit row construction feature supported in all of the
databases to do paged support so that they all optimize well.
Change-Id: Iac8703929ba2a59ee3ffa619b916d12663422887
This reverts commit 8772867855.
for uplink versions v0.25.0 through v0.30.7, there's a bug with multiplesegment upload
where the last segment is inline caused by this commit.
Change-Id: If375e186b23265586caf08991c25980e99f3cc1a
Change is adding object deletion to BeginObject request (before upload).
Now when satellite controls deletion we can move deletion before upload
to satellite. This change improves two things:
* no need for additional request to delete object before upload (need
one more change to storj/uplink)
* fix an issue with lack of permissions to upload if caveat allows only
for writing (e.g. disallow deletes but allows to write)
https://storjlabs.atlassian.net/browse/V3-3362
Change-Id: Ic453146298cdd302df290c532123731a3f99e38e
The old migration was not working. It was updateding pending (status 0)
and failed (status -1) to completed (status 100).
Change-Id: I808ff3cc692fe6c698ce26a8b411b134e67b752b
For the last few month we had no issues with order submission. I would
call it stable and now it is time to risk a lower expire time. This will
increase the database performance on the satellite and it will reduce
the delay for billing.
The long term goal is 6h but for that step we need to change graceful
exit first. At the moment storage nodes would get disuqlaified for not
transfering alle pieces in less than 6 hours.
Change-Id: I421a2c2421c5374c4e706e2338f1c2161fedc14c
paths are organized as follows:
project_id/segment_index/bucket_name/encrypted_key
so by picking parts[0] and parts[1], we were using the segment
index instead of the bucket name, causing bandwidth to be
accounted for incorrectly. additionally, we were using the
PUT action instead of the PUT_GRACEFUL_EXIT action, causing
the data to be charged incorrectly. we use PUT_REPAIR for
now because nodes won't accept uploads with PUT_GRACEFUL_EXIT
and our tables need migrations to handle rollups with it.
Change-Id: Ife2aff541222bac930c35df8fcf76e8bac5d60b2
Enable a new golangci-lint linter that has been added to the last
release. It reports a very little number of issues so they are fix it in
this commit.
Change-Id: I74fef4779c3f592aae19103fd9f70103586fe24e
Change DeleteObjectPieces for deleting the segments' pointers of an
object in a reverse order.
Last segment: L
N: total number of segments
Deleting in reverse order is: L, n-1 to 0
Deleting in reverse order makes BeginDeleteObject usable to delete
partially uploaded objects that were interrupted (e.g. upload
cancellation).
With this change, the uplink upload cancellation, can be changed to use
BeginDeleteObject to cleanup already uploaded segments without having to
retrieve orders and dial every single node which stored a piece.
Ticket: https://storjlabs.atlassian.net/browse/V3-3525
Change-Id: Ieca6fd3801c4b71671811cb5f08a99d5146928a6
A uuid.UUID is an array of bytes, and slicing it refers to the
underlying value, much like taking the address. Because range
in Go reuses the same value for every loop iteration, this means
that later iterations would overwrite earlier stored project
ids. We fix that by making a copy of the value before slicing it
for every loop iteration.
Change-Id: Iae3f11138d11a176ce360bd5af2244307c74fdad
A few variables were not renamed to the new standard piecesTotal and
piecesContentSize, so it was unclear which value was being used. These
have been updated, and some comments made more thorough.
Change-Id: I363bad4dec2a8e5c54d22c3c4cd85fc3d2b3096c
This change updates the storagenode piecestore apis to expose access to
the full piece size stored on disk. Previously we only had access to
(and only kept a cache of) the content size used for all pieces. This
was inaccurate when reporting the amount of disk space used by nodes.
We now have access to the total content size, as well as the total disk
usage, of all pieces. The pieces cache also keeps a cache of the total
piece size along with the content size.
Change-Id: I4fffe7e1257e04c46021a2e37c5adc6fe69bee55
Currently we risk losing pending bandwidth rollup writes even on a clean
shutdown. This change ensures that all pending writes are actually
written to the db when shutting down the satellite.
Change-Id: Ideab62fa9808937d3dce9585c52405d8c8a0e703
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
Currently Cockroach DB setup takes a significant amount of time.
This flattens the database setup into a single query,
which improves the test time significantly.
The migration tests still test each migration separately.
Change-Id: Iaca16f34a6af3926fa2b5ebf618f939fd59460b3
With this change RS configuration will be set on satellite. Uplink with
get RS values with BeginObject request and will use it. For backward
compatibility and to avoid super large change redundancy scheme stored
with bucket is not touched. This can be done in future.
Change-Id: Ia5f76fc10c37e2c44e4f7b8754f28eafe1f97eff
Since incoming times may be in any time zone, and we want the output
to be in UTC and for them to have 00:00:00 hours, minutes and seconds
we first convert the incoming timestamp to UTC before doing the
truncate to the day and adding a day.
Because the old code always returned a timestamp that was in the
future, this is just for efficiency.
Change-Id: Ie692d47bca8691e73852c822d5c56cf8773d99b4
Limits how many times metainfo APIs can be called per second by project ID. If limit is exceeded, the API will return Unauthorized/Too Many requests.
Limit per second and the size of the limiter cache per project are configurable, as well as whether the limiter is enabled.
Tests added/updated for the new rate_limit field in projects table.
Tests added for exceeding limits and disableing limiter.
Change-Id: Ic8ad102de3b690a475809d4f684156d5715f20fa
This change is a special case for batch processing. If in batch request
CommitSegment and CommitObject are one after another we can execute
these request as one. This will avoid current logic where we are saving
pointer for CommitSegment and later we are deleting this pointer and
saving it once again as under last segment path for CommitObject.
Change-Id: If170e78c8410f5ba5916cbff6a29b9221db9ce2e
Replace all the remaining uses of sql.DB with tagsql.DB to
fix issues with context cancellation.
Introduce tagsql.Open which helps to get rid of all tagsql.Wrap-s.
Use tagsql in cockroachkv and postgreskv.
Change-Id: I8946d203341cb85a25976896fc7881e1f704e779
Not having a skew caused an issue where:
1. Uplink calls "begin segment", where segment isn't committed to the
database.
2. Uplink stores piece X to the storage node A with timestamp 1.
3. Satellite runs garbage collection with timestamp 2.
4. Satellite sends retain request to storage node A with timestamp 2.
5. Storage node A deletes piece X, because 1 < 2.
6. Uplink calls "commit segment" with storage node A in it.
7. Download of segment fails, because A doesn't have piece X.
In production this is not an issue since the MaxTimeSkew is 72h by
default.
Change-Id: Id87ca3ddc44103dcd85d031b1367168c014b8e7b
Also added temporary types withRebind and withTagTx,
which will be later removed. Currently they help to avoid
changing the whole codebase at the same time.
Change-Id: I7f07ba8f4709a23a463bfa67464628665a05808f
warning: databases migrated to version 77 before this commit
is merged must be manually re-migrated. this should not be a
problem for anything but staging databases.
Change-Id: Ie1631c48379472352014183ee43f1465e22200f7
live accounting used to be a cache to store writes before they are picked up during
the tally iteration, after which the cache is cleared. This created a window in which
users could potentially exceed the storage limit. This PR refactors live accounting to
hold current estimations of space used per project. This should also reduce DB load
since we no longer need to query the satellite DB when checking space used for limiting.
The mechanism by which the new live accounting system works is as follows:
During the upload of any segment, the size of that segment is added to its respective
project total in live accounting. At the beginning of the tally iteration we record
the current values in live accounting as `initialLiveTotals`. At the end of the tally
iteration we again record the current totals in live accounting as `latestLiveTotals`.
The metainfo loop observer in tally allows us to get the project totals from what it
observed in metainfo DB which are stored in `tallyProjectTotals`. However, for any
particular segment uploaded during the metainfo loop, the observer may or may not
have seen it. Thus, we take half of the difference between `latestLiveTotals` and
`initialLiveTotals`, and add that to the total that was found during tally and set that
as the new live accounting total.
Initially, live accounting was storing the total stored amount across all nodes rather than
the segment size, which is inconsistent with how we record amounts stored in the project
accounting DB, so we have refactored live accounting to record segment size
Change-Id: Ie48bfdef453428fcdc180b2d781a69d58fd927fb
this commit introduces the reported_serials table. its purpose is
to allow for blind writes into it as nodes report in so that we have
minimal contention. in order to continue to accurately account for
used bandwidth, though, we cannot immediately add the settled amount.
if we did, we would have to give up on blind writes.
the table's primary key is structured precisely so that we can quickly
find expired orders and so that we maximally benefit from rocksdb
path prefix compression. we do this by rounding the expires at time
forward to the next day, effectively giving us storagenode petnames
for free. and since there's no secondary index or foreign key
constraints, this design should use significantly less space than
the current used_serials table while also reducing contention.
after inserting the orders into the table, we have a chore that
periodically consumes all of the expired orders in it and inserts
them into the existing rollups tables. this is as if we changed
the nodes to report as the order expired rather than as soon as
possible, so the belief in correctness of the refactor is higher.
since we are able to process large batches of orders (typically
a day's worth), we can use the code to maximally batch inserts into
the rollup tables to make inserts as friendly as possible to
cockroach.
Change-Id: I25d609ca2679b8331979184f16c6d46d4f74c1a6
everyone was importing it as dbx anyway. why should it be
named satellitedb? so yeah just pass the "-p dbx" flag.
Change-Id: I5efa669f4f00f196b38a9acd0d402009475a936f
Create a service for deleting pieces of storage nodes.
Currently the DeletePieces method returns after a success threshold,
completion or a timeout.
The end goal is to return when reaching the success threshold and
leaving the remaining goroutines running after DeletePieces method
returns and add a life cycle to the service that it waits for them when
it closes.
This is the first commit for ticket:
https://storjlabs.atlassian.net/browse/V3-3476
Change-Id: If740bbf57c741f880449980b8176b036dd956c7b
This reverts commit 8e242cd012.
Revert because lib/pq has known issues with context cancellation.
These issues need to be resolved before these changes can be merged.
Change-Id: I160af51dbc2d67c5449aafa406a403e5367bb555
this will allow for some nice runtime analysis down the road.
also, this allows for wrapping database handles in a way that
can interact with these contexts
requires https://review.dev.storj.io/c/storj/dbx/+/514
Change-Id: Ib087b7cd73296dd2c1e0331314da34d861f61d2b
this allows for setting $STORJ_METAINFO_POSTGRESQL_USE_ALT=yes if you
want to use the cockroachkv implementation for metainfo against postgres
Change-Id: I0c9458c83fd67ee63ef4a78351e64a80a0647408
the hope is that it is mostly interfering with itself, so this
will make it not do that (well, N api servers, but hopefully
that's not enough to cause it to have issues).
Change-Id: Ifd0c9e6617457785ab25fe5b714d8556cdc8e2d3
When an uplink requests an upload or download from the satellite we are trackig the
allocated bandwidth twice. The value in bucket_bandwidth_rollups is used
for project limits but the value in storagenode_bandwidth_rollups is not
used at all. We can increase the performance by removing it. Uplinks
will get a faster response from the satellite.
Change-Id: Icccd41f94107ef34668f30f99bf5f728c384b07e
any database error doesn't mean the order wasn't found. for example
in cockroach it may say that the transaction is aborted. then what?
maybe we get big old row level deadlocks like we've observed? so
instead explicitly check for ErrNoRows to reject the order and bail
out otherwise. the surrounding logic will give it a retry.
Change-Id: I6e1f8f6e6a6def3e45b44f5088cbdc158e1098e4
Add a back-pressure mechanism to the satellite metainfo
DeleteObjectPieces method for returning once the 75% of successful
deleted pieces is reached.
Change-Id: Ia38df49fba5838f0605c40a77cfff8e3442cb5b0
With the new storage node downtime tracking feature, we need remove current uptime reputation configs: UptimeReputationAlpha, UptimeReputationBeta, and
UptimeReputationDQ. This is the first step of removing the uptime
reputation columns from satellitedb
Change-Id: Ie8fab13295dbf545e33aeda0c4306cda4ba54e36
The DeleteObjectPieces should print out the warning on closing the
connections only if there was an error.
Change-Id: If3d7ab256d8508c08388c1f22c7dd1eb819d2509
The DeleteObjectPieces must close the storage node client once it has
finished deleting its pieces.
Change-Id: I08eb8af8e4215d77d59b52f5055211b918374ab4
turns out portable sed is hard: it has to work with both
linux and bsd sed, etc. instead, use a really really basic
bash script and a temporary file. this should be much less
likely to cause issues on a wide range of machines.
Change-Id: Ia759789fb52aa1ee3361426bb6c02ed4eac3d23a
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
I also fleshed out consoledb_test.go to do actual inserts and gets to
make sure things were working correctly.
Change-Id: I089bf4c776d15dc8578080e26760bd6dff4beec9
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
Change-Id: I22b850ce5859fa07d13bf475be5140e6bde95b8a
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The WithTx
helper functions in dbutil and dbx do that automatically. I am changing
this code to use those helpers instead.
Change-Id: Iaf492af35471931125f2b7365aa4338f44154881
DeleteObjectPieces must not call overlay cache KnownReliable method with
an empty list of node IDs for avoiding to log a useless noisy warning.
Change-Id: Ibe2a34f2913f003d3ba020f9764c1369fa63123b
Move tests for old Metainfo API to separate file. Metainfo tests file is
large enough and in future it will be easier to remove old tests.
Change-Id: I9421907ef015a6dfa65f4de6ef01b2d2c8baa7df
Use the helper function IsRPC of the err2 package rather than checking
if an error is of a specific RPC status code with an 'if' conditional.
Change-Id: Ibe89d6c2d836307c3112a6d7cc6bf95f0f985fd2
Disqualifies a node when the node fails to complete a graceful
exit.
Adds a new DisqualifyNode method to the overlay cache, since there
wasn't an existing method to disqualify a node but do nothing else
to its stats.
Adds checks to existing tests to make sure that a storage node that
fails a graceful exit is marked as disqualified in the overlay
cache.
https: //storjlabs.atlassian.net/browse/V3-3342
Change-Id: I4d554a519ab59db31ad3b8e28764c8683a6e3888
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
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
Change-Id: I660540885a0784fae844cf99376d1537e208fa69
overlay.GetOfflineNodesLimited
We only care about node ID, address, and last contact success/failure
from the downtime service, so the overlay should only return these
values for the downtime-specific queries.
Change-Id: I08a6ecfdd2a12b82cae62e87d6adeab53975bfce
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
Change-Id: Icd3da71448a84c582c6afdc6b52d1f345fe9469f
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
Change-Id: Ibaadd2c8540ba5c8cccd6ecbf529017ab98b78ca
Transactions in our code that might need to work against CockroachDB
need to be retried in the event of a retryable error. The transaction
helper functions in dbutil do that automatically. I am changing this
code to use those helpers instead.
Change-Id: Id24906f5f3ae83245dabb218e1f70e0bcb3b417a
Remove starting up messages from peers. We expect all of them to start,
if they don't, then they should return an error why they don't start.
The only informative message is when a service is disabled.
When doing initial database setup then each migration step isn't
informative, hence print only a single line with the final version.
Also use shorter log scopes.
Change-Id: Ic8b61411df2eeae2a36d600a0c2fbc97a84a5b93
