Design: https://gist.github.com/klauspost/025c09b48ed4a1293c917cecfabdf21c
Gist of improvements:
* Cross-server caching and listing will use the same data across servers and requests.
* Lists can be arbitrarily resumed at a constant speed.
* Metadata for all files scanned is stored for streaming retrieval.
* The existing bloom filters controlled by the crawler is used for validating caches.
* Concurrent requests for the same data (or parts of it) will not spawn additional walkers.
* Listing a subdirectory of an existing recursive cache will use the cache.
* All listing operations are fully streamable so the number of objects in a bucket no
longer dictates the amount of memory.
* Listings can be handled by any server within the cluster.
* Caches are cleaned up when out of date or superseded by a more recent one.
- select lockers which are non-local and online to have
affinity towards remote servers for lock contention
- optimize lock retry interval to avoid sending too many
messages during lock contention, reduces average CPU
usage as well
- if bucket is not set, when deleteObject fails make sure
setPutObjHeaders() honors lifecycle only if bucket name
is set.
- fix top locks to list out always the oldest lockers always,
avoid getting bogged down into map's unordered nature.
This is to ensure that Go contexts work properly, after some
interesting experiments I found that Go net/http doesn't
cancel the context when Body is non-zero and hasn't been
read till EOF.
The following gist explains this, this can lead to pile up
of go-routines on the server which will never be canceled
and will die at a really later point in time, which can
simply overwhelm the server.
https://gist.github.com/harshavardhana/c51dcfd055780eaeb71db54f9c589150
To avoid this refactor the locking such that we take locks after we
have started reading from the body and only take locks when needed.
Also, remove contextReader as it's not useful, doesn't work as expected
context is not canceled until the body reaches EOF so there is no point
in wrapping it with context and putting a `select {` on it which
can unnecessarily increase the CPU overhead.
We will still use the context to cancel the lockers etc.
Additional simplification in the locker code to avoid timers
as re-using them is a complicated ordeal avoid them in
the hot path, since locking is very common this may avoid
lots of allocations.
MaxConnsPerHost can potentially hang a call without any
way to timeout, we do not need this setting for our proxy
and gateway implementations instead IdleConn settings are
good enough.
Also ensure to use NewRequestWithContext and make sure to
take the disks offline only for network errors.
Fixes#10304
Add context to all (non-trivial) calls to the storage layer.
Contexts are propagated through the REST client.
- `context.TODO()` is left in place for the places where it needs to be added to the caller.
- `endWalkCh` could probably be removed from the walkers, but no changes so far.
The "dangerous" part is that now a caller disconnecting *will* propagate down, so a
"delete" operation will now be interrupted. In some cases we might want to disconnect
this functionality so the operation completes if it has started, leaving the system in a cleaner state.
- delete-marker should be created on a suspended bucket as `null`
- delete-marker should delete any pre-existing `null` versioned
object and create an entry `null`
when source and destination are same and versioning is enabled
on the destination bucket - we do not need to re-create the entire
object once again to optimize on space utilization.
Cases this PR is not supporting
- any pre-existing legacy object will not
be preserved in this manner, meaning a new
dataDir will be created.
- key-rotation and storage class changes
of course will never re-use the dataDir
With reduced parity our write quorum should be same
as read quorum, but code was still assuming
```
readQuorum+1
```
In all situations which is not necessary.
Bonus fix during versioning merge one of the PR was missing
the offline/online disk count fix from #9801 port it correctly
over to the master branch from release.
Additionally, add versionID support for MRF
Fixes#9910Fixes#9931
Just like GET/DELETE APIs it is possible to preserve
client supplied versionId's, of course the versionIds
have to be uuid, if an existing versionId is found
it is overwritten if no object locking policies
are found.
- PUT /bucketname/objectname?versionId=<id>
- POST /bucketname/objectname?uploads=&versionId=<id>
- PUT /bucketname/objectname?verisonId=<id> (with x-amz-copy-source)
PutObject on multiple-zone with versioning would not
overwrite the correct location of the object if the
object has delete marker, leading to duplicate objects
on two zones.
This PR fixes by adding affinity towards delete marker
when GetObjectInfo() returns error, use the zone index
which has the delete marker.
- Implement a new xl.json 2.0.0 format to support,
this moves the entire marshaling logic to POSIX
layer, top layer always consumes a common FileInfo
construct which simplifies the metadata reads.
- Implement list object versions
- Migrate to siphash from crchash for new deployments
for object placements.
Fixes#2111
CopyObject was not correctly figuring out the correct
destination object location and would end up creating
duplicate objects on two different zones, reproduced
by doing encryption based key rotation.
If the requested server is part of the set this will always read
from the local disk, even if the disk contains a parity shard.
In default setup there is a 50% chance that at least
one shard that otherwise would have been fetched remotely
will be read locally instead.
It basically trades RPC call overhead for reed-solomon.
On distributed localhost this seems to be fairly break-even,
with a very small gain in throughput and latency.
However on networked servers this should be a bigger
1MB objects, before:
```
Operation: GET. Concurrency: 32. Hosts: 4.
Requests considered: 76257:
* Avg: 25ms 50%: 24ms 90%: 32ms 99%: 42ms Fastest: 7ms Slowest: 67ms
* First Byte: Average: 23ms, Median: 22ms, Best: 5ms, Worst: 65ms
Throughput:
* Average: 1213.68 MiB/s, 1272.63 obj/s (59.948s, starting 14:45:44 CEST)
```
After:
```
Operation: GET. Concurrency: 32. Hosts: 4.
Requests considered: 78845:
* Avg: 24ms 50%: 24ms 90%: 31ms 99%: 39ms Fastest: 8ms Slowest: 62ms
* First Byte: Average: 22ms, Median: 21ms, Best: 6ms, Worst: 57ms
Throughput:
* Average: 1255.11 MiB/s, 1316.08 obj/s (59.938s, starting 14:43:58 CEST)
```
Bonus fix: Only ask for heal once on an object.
By monitoring PUT/DELETE and heal operations it is possible
to track changed paths and keep a bloom filter for this data.
This can help prioritize paths to scan. The bloom filter can identify
paths that have not changed, and the few collisions will only result
in a marginal extra workload. This can be implemented on either a
bucket+(1 prefix level) with reasonable performance.
The bloom filter is set to have a false positive rate at 1% at 1M
entries. A bloom table of this size is about ~2500 bytes when serialized.
To not force a full scan of all paths that have changed cycle bloom
filters would need to be kept, so we guarantee that dirty paths have
been scanned within cycle runs. Until cycle bloom filters have been
collected all paths are considered dirty.
global WORM mode is a complex piece for which
the time has passed, with the advent of S3 compatible
object locking and retention implementation global
WORM is sort of deprecated, this has been mentioned
in our documentation for some time, now the time
has come for this to go.
Too many deployments come up with an odd number
of hosts or drives, to facilitate even distribution
among those setups allow for odd and prime numbers
based packs.
Bulk delete API was using cleanupObjectsBulk() which calls posix
listing and delete API to remove objects internal files in the
backend (xl.json and parts) one by one.
Add DeletePrefixes in the storage API to remove the content
of a directory in a single call.
Also use a remove goroutine for each disk to accelerate removal.
- Remove the requirement to honor storage class for deletes
- Improve `posix.DeleteFileBulk` code to Stat the volumeDir
only once per call, rather than for all object paths.
This is a precursor change before versioning,
removes/deprecates the requirement of remembering
partName and partETag which are not useful after
a multipart transaction has finished.
This PR reduces the overall size of the backend
JSON for large file uploads.
multi-delete API failed with write quorum errors
under following situations
- list of files requested for delete doesn't exist
anymore can lead to quorum errors and failure
- due to usage of query param for paths, for really
long paths MinIO server rejects these requests as
malformed as unexpected.
This was reproduced with warp
object lock config is enabled for a bucket.
Creating a bucket with object lock configuration
enabled does not automatically cause WORM protection
to be applied. PUT operation needs to specifically
request object locking or bucket has to have default
retention settings configured.
Fixes regression introduced in #8657
This ensures that we can update the
- .minio.sys is updated for accounting/data usage purposes
- .minio.sys is updated to indicate if backend is encrypted
or not.
Klaus Post