Object names of directory objects qualified for ExpiredObjectAllVersions
must be encoded appropriately before calling on deletePrefix on their
erasure set.
e.g., a directory object and regular objects with overlapping prefixes
could lead to the expiration of regular objects, which is not the
intention of ILM.
```
bucket/dir/ ---> directory object
bucket/dir/obj-1
```
When `bucket/dir/` qualifies for expiration, the current implementation would
remove regular objects under the prefix `bucket/dir/`, in this case,
`bucket/dir/obj-1`.
This commit changes how MinIO generates the object encryption key (OEK)
when encrypting an object using server-side encryption.
This change is fully backwards compatible. Now, MinIO generates
the OEK as following:
```
Nonce = RANDOM(32) // generate 256 bit random value
OEK = HMAC-SHA256(EK, Context || Nonce)
```
Before, the OEK was computed as following:
```
Nonce = RANDOM(32) // generate 256 bit random value
OEK = SHA256(EK || Nonce)
```
The new scheme does not technically fix a security issue but
uses a more familiar scheme. The only requirement for the
OEK generation function is that it produces a (pseudo)random value
for every pair (`EK`,`Nonce`) as long as no `EK`-`Nonce` combination
is repeated. This prevents a faulty PRNG from repeating or generating
a "bad" key.
The previous scheme guarantees that the `OEK` is a (pseudo)random
value given that no pair (`EK`,`Nonce`) repeats under the assumption
that SHA256 is indistinguable from a random oracle.
The new scheme guarantees that the `OEK` is a (pseudo)random value
given that no pair (`EK`, `Nonce`) repeats under the assumption that
SHA256's underlying compression function is a PRF/PRP.
While the later is a weaker assumption, and therefore, less likely
to be false, both are considered true. SHA256 is believed to be
indistinguable from a random oracle AND its compression function
is assumed to be a PRF/PRP.
As far as the OEK generating is concerned, the OS random number
generator is not required to be pseudo-random but just non-repeating.
Apart from being more compatible to standard definitions and
descriptions for how to generate crypto. keys, this change does not
have any impact of the actual security of the OEK key generation.
Signed-off-by: Andreas Auernhammer <github@aead.dev>
avoids error during upgrades such as
```
API: SYSTEM()
Time: 19:19:22 UTC 03/18/2024
DeploymentID: 24e4b574-b28d-4e94-9bfa-03c363a600c2
Error: Invalid api configuration: found invalid keys (expiry_workers=100 ) for 'api' sub-system, use 'mc admin config reset myminio api' to fix invalid keys (*fmt.wrapError)
11: internal/logger/logger.go:260:logger.LogIf()
...
```
we were prematurely not writing 4k pages while we
could have due to the fact that most buffers would
be multiples of 4k upto some number and there shall
be some remainder.
We only need to write the remainder without O_DIRECT.
at scale customers might start with failed drives,
causing skew in the overall usage ratio per EC set.
make this configurable such that customers can turn
this off as needed depending on how comfortable they
are.
Cosmetic change, but breaks up a big code block and will make a goroutine
dumps of streams are more readable, so it is clearer what each goroutine is doing.
When an object qualifies for both tiering and expiration rules and is
past its expiration date, it should be expired without requiring to tier
it, even when tiering event occurs before expiration.
configure batch size to send audit/logger events
in batches instead of sending one event per connection.
this is mainly to optimize the number of requests
we make to webhook endpoint.
simplify audit webhook worker model
fixes couple of bugs like
- ping(ctx) was creating a logger without updating
number of workers leading to incorrect nWorkers
scaling, causing an additional worker that is not
tracked properly.
- h.logCh <- entry could potentially hang for when
the queue is full on heavily loaded systems.
there can be a sudden spike in tiny allocations,
due to too much auditing being done, also don't hang
on the
```
h.logCh <- entry
```
after initializing workers if you do not have a way to
dequeue for some reason.
Remove api.expiration_workers config setting which was inadvertently left behind. Per review comment
https://github.com/minio/minio/pull/18926, expiration_workers can be configured via ilm.expiration_workers.
- Use a shared worker pool for all ILM expiry tasks
- Free version cleanup executes in a separate goroutine
- Add a free version only if removing the remote object fails
- Add ILM expiry metrics to the node namespace
- Move tier journal tasks to expiryState
- Remove unused on-disk journal for tiered objects pending deletion
- Distribute expiry tasks across workers such that the expiry of versions of
the same object serialized
- Ability to resize worker pool without server restart
- Make scaling down of expiryState workers' concurrency safe; Thanks
@klauspost
- Add error logs when expiryState and transition state are not
initialized (yet)
* metrics: Add missed tier journal entry tasks
* Initialize the ILM worker pool after the object layer
With this commit, MinIO generates root credentials automatically
and deterministically if:
- No root credentials have been set.
- A KMS (KES) is configured.
- API access for the root credentials is disabled (lockdown mode).
Before, MinIO defaults to `minioadmin` for both the access and
secret keys. Now, MinIO generates unique root credentials
automatically on startup using the KMS.
Therefore, it uses the KMS HMAC function to generate pseudo-random
values. These values never change as long as the KMS key remains
the same, and the KMS key must continue to exist since all IAM data
is encrypted with it.
Backward compatibility:
This commit should not cause existing deployments to break. It only
changes the root credentials of deployments that have a KMS configured
(KES, not a static key) but have not set any admin credentials. Such
implementations should be rare or not exist at all.
Even if the worst case would be updating root credentials in mc
or other clients used to administer the cluster. Root credentials
are anyway not intended for regular S3 operations.
Signed-off-by: Andreas Auernhammer <github@aead.dev>
just like client-conn-read-deadline, added a new flag that does
client-conn-write-deadline as well.
Both are not configured by default, since we do not yet know
what is the right value. Allow this to be configurable if needed.
we should do this to ensure that we focus on
data healing as primary focus, fixing metadata
as part of healing must be done but making
data available is the main focus.
the main reason is metadata inconsistencies can
cause data availability issues, which must be
avoided at all cost.
will be bringing in an additional healing mechanism
that involves "metadata-only" heal, for now we do
not expect to have these checks.
continuation of #19154
Bonus: add a pro-active healthcheck to perform a connection
Streams can return errors if the cancelation is picked up before the response
stream close is picked up. Under extreme load, this could lead to missing
responses.
Send server mux ack async so a blocked send cannot block newMuxStream
call. Stream will not progress until mux has been acked.
in k8s things really do come online very asynchronously,
we need to use implementation that allows this randomness.
To facilitate this move WriteAll() as part of the
websocket layer instead.
Bonus: avoid instances of dnscache usage on k8s
Currently, if one object tag matches with one lifecycle tag filter, ILM
will select it, however, this is wrong. All the Tag filters in the
lifecycle document should be satisfied.
Currently, we read from `/proc/diskstats` which is found to be
un-reliable in k8s environments. We can read from `sysfs` instead.
Also, cache the latest drive io stats to find the diff and update
the metrics.
* Remove lock for cached operations.
* Rename "Relax" to `ReturnLastGood`.
* Add `CacheError` to allow caching values even on errors.
* Add NoWait that will return current value with async fetching if within 2xTTL.
* Make benchmark somewhat representative.
```
Before: BenchmarkCache-12 16408370 63.12 ns/op 0 B/op
After: BenchmarkCache-12 428282187 2.789 ns/op 0 B/op
```
* Remove `storageRESTClient.scanning`. Nonsensical - RPC clients will not have any idea about scanning.
* Always fetch remote diskinfo metrics and cache them. Seems most calls are requesting metrics.
* Do async fetching of usage caches.
If network conditions have filled the output queue before a reconnect happens blocked sends could stop reconnects from happening. In short `respMu` would be held for a mux client while sending - if the queue is full this will never get released and closing the mux client will hang.
A) Use the mux client context instead of connection context for sends, so sends are unblocked when the mux client is canceled.
B) Use a `TryLock` on "close" and cancel the request if we cannot get the lock at once. This will unblock any attempts to send.
Add a new function logger.Event() to send the log to Console and
http/kafka log webhooks. This will include some internal events such as
disk healing and rebalance/decommissioning
Bonus: enable audit alerts for object versions
beyond the configured value, default is '100'
versions per object beyond which scanner will
alert for each such objects.
Fix reported races that are actually synchronized by network calls.
But this should add some extra safety for untimely disconnects.
Race reported:
```
WARNING: DATA RACE
Read at 0x00c00171c9c0 by goroutine 214:
github.com/minio/minio/internal/grid.(*muxClient).addResponse()
e:/gopath/src/github.com/minio/minio/internal/grid/muxclient.go:519 +0x111
github.com/minio/minio/internal/grid.(*muxClient).error()
e:/gopath/src/github.com/minio/minio/internal/grid/muxclient.go:470 +0x21d
github.com/minio/minio/internal/grid.(*Connection).handleDisconnectClientMux()
e:/gopath/src/github.com/minio/minio/internal/grid/connection.go:1391 +0x15b
github.com/minio/minio/internal/grid.(*Connection).handleMsg()
e:/gopath/src/github.com/minio/minio/internal/grid/connection.go:1190 +0x1ab
github.com/minio/minio/internal/grid.(*Connection).handleMessages.func1()
e:/gopath/src/github.com/minio/minio/internal/grid/connection.go:981 +0x610
Previous write at 0x00c00171c9c0 by goroutine 1081:
github.com/minio/minio/internal/grid.(*muxClient).roundtrip()
e:/gopath/src/github.com/minio/minio/internal/grid/muxclient.go:94 +0x324
github.com/minio/minio/internal/grid.(*muxClient).traceRoundtrip()
e:/gopath/src/github.com/minio/minio/internal/grid/trace.go:74 +0x10e4
github.com/minio/minio/internal/grid.(*Subroute).Request()
e:/gopath/src/github.com/minio/minio/internal/grid/connection.go:366 +0x230
github.com/minio/minio/internal/grid.(*SingleHandler[go.shape.*github.com/minio/minio/cmd.DiskInfoOptions,go.shape.*github.com/minio/minio/cmd.DiskInfo]).Call()
e:/gopath/src/github.com/minio/minio/internal/grid/handlers.go:554 +0x3fd
github.com/minio/minio/cmd.(*storageRESTClient).DiskInfo()
e:/gopath/src/github.com/minio/minio/cmd/storage-rest-client.go:314 +0x270
github.com/minio/minio/cmd.erasureObjects.getOnlineDisksWithHealingAndInfo.func1()
e:/gopath/src/github.com/minio/minio/cmd/erasure.go:293 +0x171
```
This read will always happen after the write, since there is a network call in between.
However a disconnect could come in while we are setting up the call, so we protect against that with extra checks.
- bucket metadata does not need to look for legacy things
anymore if b.Created is non-zero
- stagger bucket metadata loads across lots of nodes to
avoid the current thundering herd problem.
- Remove deadlines for RenameData, RenameFile - these
calls should not ever be timed out and should wait
until completion or wait for client timeout. Do not
choose timeouts for applications during the WRITE phase.
- increase R/W buffer size, increase maxMergeMessages to 30
We have observed cases where a blocked stream will block for cancellations.
This happens when response channel is blocked and we want to push an error.
This will have the response mutex locked, which will prevent all other operations until upstream is unblocked.
Make this behavior non-blocking and if blocked spawn a goroutine that will send the response and close the output.
Still a lot of "dancing". Added a test for this and reviewed.
Depending on when the context cancelation is picked up the handler may return and close the channel before `SubscribeJSON` returns, causing:
```
Feb 05 17:12:00 s3-us-node11 minio[3973657]: panic: send on closed channel
Feb 05 17:12:00 s3-us-node11 minio[3973657]: goroutine 378007076 [running]:
Feb 05 17:12:00 s3-us-node11 minio[3973657]: github.com/minio/minio/internal/pubsub.(*PubSub[...]).SubscribeJSON.func1()
Feb 05 17:12:00 s3-us-node11 minio[3973657]: github.com/minio/minio/internal/pubsub/pubsub.go:139 +0x12d
Feb 05 17:12:00 s3-us-node11 minio[3973657]: created by github.com/minio/minio/internal/pubsub.(*PubSub[...]).SubscribeJSON in goroutine 378010884
Feb 05 17:12:00 s3-us-node11 minio[3973657]: github.com/minio/minio/internal/pubsub/pubsub.go:124 +0x352
```
Wait explicitly for the goroutine to exit.
Bonus: Listen for doneCh when sending to not risk getting blocked there is channel isn't being emptied.
Krishnan Parthasarathi
Anis Eleuch
Andreas Auernhammer
Harshavardhana
jiuker
Klaus Post
huajin tong
Aditya Manthramurthy
Praveen raj Mani
Shubhendu
Taran Pelkey