For large objects taking more than '3 minutes' response
times in a single PUT operation can timeout prematurely
as 'ResponseHeader' timeout hits for 3 minutes. Avoid
this by keeping the connection active during CreateFile
phase.
For large objects taking more than '3 minutes' response
times in a single PUT operation can timeout prematurely
as 'ResponseHeader' timeout hits for 3 minutes. Avoid
this by keeping the connection active during CreateFile
phase.
* Provide information on *actively* healing, buckets healed/queued, objects healed/failed.
* Add concurrent healing of multiple sets (typically on startup).
* Add bucket level resume, so restarts will only heal non-healed buckets.
* Print summary after healing a disk is done.
also additionally make sure errors during deserializer closes
the reader with right error type such that Write() end
actually see the final error, this avoids a waitGroup usage
and waiting.
additionally simply timedValue to have RWMutex
to avoid concurrent calls to DiskInfo() getting
serialized, this has an effect on all calls that
use GetDiskInfo() on the same disks.
Such as getOnlineDisks, getOnlineDisksWithoutHealing
This PR refactors the way we use buffers for O_DIRECT and
to re-use those buffers for messagepack reader writer.
After some extensive benchmarking found that not all objects
have this benefit, and only objects smaller than 64KiB see
this benefit overall.
Benefits are seen from almost all objects from
1KiB - 32KiB
Beyond this no objects see benefit with bulk call approach
as the latency of bytes sent over the wire v/s streaming
content directly from disk negate each other with no
remarkable benefits.
All other optimizations include reuse of msgp.Reader,
msgp.Writer using sync.Pool's for all internode calls.
Rewrite parentIsObject() function. Currently if a client uploads
a/b/c/d, we always check if c, b, a are actual objects or not.
The new code will check with the reverse order and quickly quit if
the segment doesn't exist.
So if a, b, c in 'a/b/c' does not exist in the first place, then returns
false quickly.
The only purpose of check-dir flag in
ReadVersion is to return 404 when
an object has xl.meta but without data.
This is causing an extract call to the disk
which can be penalizing in case of busy system
where disks receive many concurrent access.
optimization mainly to avoid listing the entire
`.minio.sys/buckets/.minio.sys` directory, this
can get really huge and comes in the way of startup
routines, contents inside `.minio.sys/buckets/.minio.sys`
are rather transient and not necessary to be healed.
fixes a regression introduced in #10859, due
to the error returned by rest.Client being typed
i.e *rest.NetworkError - IsNetworkHostDown function
didn't work as expected to detect network issues.
This in-turn aggravated the situations when nodes
are disconnected leading to performance loss.
This will make the health check clients 'silent'.
Use `IsNetworkOrHostDown` determine if network is ok so it mimics the functionality in the actual client.
Similar to #10775 for fewer memory allocations, since we use
getOnlineDisks() extensively for listing we should optimize it
further.
Additionally, remove all unused walkers from the storage layer
WriteAll saw 127GB allocs in a 5 minute timeframe for 4MiB buffers
used by `io.CopyBuffer` even if they are pooled.
Since all writers appear to write byte buffers, just send those
instead and write directly. The files are opened through the `os`
package so they have no special properties anyway.
This removes the alloc and copy for each operation.
REST sends content length so a precise alloc can be made.
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.
only newly replaced drives get the new `format.json`,
this avoids disks reloading their in-memory reference
format, ensures that drives are online without
reloading the in-memory reference format.
keeping reference format in-tact means UUIDs
never change once they are formatted.
connect disks pre-emptively upon startup, to ensure we have
enough disks are connected at startup rather than wait
for them.
we need to do this to avoid long wait times for server to
be online when we have servers come up in rolling upgrade
fashion
`mc admin info` on busy setups will not move HDD
heads unnecessarily for repeated calls, provides
a better responsiveness for the call overall.
Bonus change allow listTolerancePerSet be N-1
for good entries, to avoid skipping entries
for some reason one of the disk went offline.
add a hint on the disk to allow for tracking fresh disk
being healed, to allow for restartable heals, and also
use this as a way to track and remove disks.
There are more pending changes where we should move
all the disk formatting logic to backend drives, this
PR doesn't deal with this refactor instead makes it
easier to track healing in the future.
Healing was not working correctly in the distributed mode because
errFileVersionNotFound was not properly converted in storage rest
client.
Besides, fixing the healing delete marker is not working as expected.
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.
this is to detect situations of corruption disk
format etc errors quickly and keep the disk online
in such scenarios for requests to fail appropriately.
- admin info node offline check is now quicker
- admin info now doesn't duplicate the code
across doing the same checks for disks
- rely on StorageInfo to return appropriate errors
instead of calling locally.
- diskID checks now return proper errors when
disk not found v/s format.json missing.
- add more disk states for more clarity on the
underlying disk errors.
Without instantiating a new rest client we can
have a recursive error which can lead to
healthcheck returning always offline, this can
prematurely take the servers offline.
- Add changes to ensure remote disks are not
incorrectly taken online if their order has
changed or are incorrect disks.
- Bring changes to peer to detect disconnection
with separate Health handler, to avoid a
rather expensive call GetLocakDiskIDs()
- Follow up on the same changes for Lockers
as well
- 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
GetDiskID() in storage rest client does not really issue a REST request
to the remote disk, but returns an in-memory value instead.
However, GetDiskID() should return an error when format.json is not
found or for other similar issues (unmounted disks, etc..)
GetDiskID() is only called when formatting disks and getting storage
informatio, hence this commit should not have a performance degradation.
Shuffling arguments that we pass to MinIO server are supported. However,
when that happens, Prometheus returns wrong information about disks usage
and online/offline status.
The commit fixes the issue by avoiding relying on xl.endpoints since
it is not ordered.
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.
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
instead perform a liveness check call to
verify if server is online and print relevant
errors.
Also introduce a StorageErr string error type
instead of errors.New() deprecate usage of
VerifyFileError, DeleteFileError for gob,
change in datastructure also requires bump in
storage REST version to v13.
Fixes#8811
When formatting a set validate if a host failure will likely lead to data loss.
While we don't know what config will be set in the future
evaluate to our best knowledge, assuming default settings.
Admin data usage info API returns the following
(Only FS & XL, for now)
- Number of buckets
- Number of objects
- The total size of objects
- Objects histogram
- Bucket sizes
This PR implements locking from a global entity into
a more localized set level entity, allowing for locks
to be held only on the resources which are writing
to a collection of disks rather than a global level.
In this process this PR also removes the top-level
limit of 32 nodes to an unlimited number of nodes. This
is a precursor change before bring in bucket expansion.
This PR adds code to appropriately handle versioning issues
that come up quite constantly across our API changes. Currently
we were also routing our requests wrong which sort of made it
harder to write a consistent error handling code to appropriately
reject or honor requests.
This PR potentially fixes issues
- old mc is used against new minio release which is incompatible
returns an appropriate for client action.
- any older servers talking to each other, report appropriate error
- incompatible peer servers should report error and reject the calls
with appropriate error
- Heal if the part.1 is truncated from its original size
- Heal if the part.1 fails while being verified in between
- Heal if the part.1 fails while being at a certain offset
Other cleanups include make sure to flush the HTTP responses
properly from storage-rest-server, avoid using 'defer' to
improve call latency. 'defer' incurs latency avoid them
in our hot-paths such as storage-rest handlers.
Fixes#8319
errors.errorString() cannot be marshalled by gob
encoder, so using a slice of []error would fail
to be encoded. This leads to no errors being
generated instead gob.Decoder on the storage-client
would see an io.EOF
To avoid such bugs introduce a typed error for
handling such translations and register this type
for gob encoding support.
VerifyFile in the distributed setup does not work with
the non streaming highway hash. The reason is that the
internode mux router did not expect `storageRESTBitrotHash`
parameter.
This avoids a network call, also fixes an issue
when empty paths are passed the underlying call
fails with "405 Method Not Allowed".
This is reproducible when you are deleting a
non-existent object.
Fixes#8083
In situations such as when client uploading data,
prematurely disconnects from server such as pressing
ctrl-c before uploading all the data. Under this
situation in distributed setup we prematurely
disconnect disks causing a reconnect loop. This has
an adverse affect we end up leaving a lot of files
in temporary location which ideally should have been
cleaned up when Put() prematurely fails.
This is also a regression which got introduced in #7610
posix.VerifyFile() doesn't know how to check if a file
is corrupted if that file is empty. We do have the part
size in xl.json so we pass it to VerifyFile to return
an error so healing empty parts can work properly.
Consider errors returned by httpClient.Do() as network errors. This is because
the http clients returns different types of errors and it is hard to catch
all the error types.
With these changes we are now able to peak performances
for all Write() operations across disks HDD and NVMe.
Also adds readahead for disk reads, which also increases
performance for reads by 3x.
Bulk delete at storage level in Multiple Delete Objects API
In order to accelerate bulk delete in Multiple Delete objects API,
a new bulk delete is introduced in storage layer, which will accept
a list of objects to delete rather than only one. Consequently,
a new API is also need to be added to Object API.
Other listing optimizations include
- remove double sorting while filtering object entries
- improve error message when upload-id is not in quorum
- use jsoniter for full unmarshal json, instead of gjson
- remove unused code
In distributed mode, use REST API to acquire and manage locks instead
of RPC.
RPC has been completely removed from MinIO source.
Since we are moving from RPC to REST, we cannot use rolling upgrades as the
nodes that have not yet been upgraded cannot talk to the ones that have
been upgraded.
We expect all minio processes on all nodes to be stopped and then the
upgrade process to be completed.
Also force http1.1 for inter-node communication
Before this commit, nodes wait indefinitely without showing any
indicate error message when a node is started with different access
and secret keys.
This PR will show '401 Unauthorized' in this case.
This commit increases storage REST requests to 5 minutes, this includes
the opening TCP connection, and sending/receiving data. This will reduce
clients receiving errors when the server is under high load.
