Commit graph

3325 commits

Author SHA1 Message Date
Krishnan Parthasarathi c33d1b8ee6 Vendorize minio/dsync for server-side read lock (#2484)
- Prevention of stale lock accumulation.
- Removal of dead code.
2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi 1f67c18222 Lock/Rlock rpc reply was incorrect (#2479) 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi a4691611a7 Move initGracefulShutdown before objectLayer initialization (#2468) 2016-09-13 21:18:30 -07:00
awwalker 7c7eb1475d splitNetPath: Add support for windows paths including volumeNames e.g ip:C:\network\path 2016-09-13 21:18:30 -07:00
Harshavardhana 0bce3d6d63 Fix web handler initialize with distributed lazy init. 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi 804d91ef61 storage/rpc-client: Reconnect on network disconnect (#2436) 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi 229600ce9b Implement RLock, RUnlock rpc handlers (#2437)
This would make it simplify dsync.RWMutex's algorithm to acquire a
distributed read lock and be tolerant to N/2-1 failures.
2016-09-13 21:18:30 -07:00
Harshavardhana 43098df9d8 rpc: Re-factor ReadFile behavior client <--> server.
Current code did not marshal/unmarshal buffers properly from
the server, The reason being buffers have to be allocated and
sent properly back to client to be consumable.
2016-09-13 21:18:30 -07:00
Harshavardhana 6908a0dcd4 Extract rpc server wrapped errors and translate to storage error. 2016-09-13 21:18:30 -07:00
Harshavardhana cae5761f16 rpc/client: Add missing rpcTokens for each rpc calls. 2016-09-13 21:18:30 -07:00
Harshavardhana 83074ed57e Add missing rpc-client.go - missing in previous commit. 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi 64a7e6992e Vendorize rpc reconnect changes from minio/dsync (#2405) 2016-09-13 21:18:30 -07:00
Harshavardhana 61af764f8a Add rpc layer authentication. 2016-09-13 21:18:30 -07:00
Harshavardhana b4172ad3c8 Rename rpc-{client,server} storage-rpc-{client,server} 2016-09-13 21:18:30 -07:00
Harshavardhana 4917038f55 Move the ObjectAPI() resource to be beginning of each handlers.
We should return back proper errors so that the clients can
retry until server has been initialized.
2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi b0f3f94163 unify single-node and distributed namespace locking (#2401) 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi b7c169d71d object layer initialization using distributed locking (#2397)
* vendorized latest minio/dsync

* wip - object layer initialization using distributed locking
2016-09-13 21:18:30 -07:00
Frank 3939c75345 Added distributed RWMutex (#2369) 2016-09-13 21:18:30 -07:00
Krishnan Parthasarathi e55926e8cf distribute: Make server work with multiple remote disks
This change initializes rpc servers associated with disks that are
local. It makes object layer initialization on demand, namely on the
first request to the object layer.

Also adds lock RPC service vendorized minio/dsync
2016-09-13 21:18:30 -07:00
Anis Elleuch f82f535509 Fix possible race in shutdown callbacks process and simplify shuttting down profiler (#2684) 2016-09-13 18:43:45 -07:00
Anis Elleuch 51e337228e Avoid hiding disk errors in some cases in FS Shutdown (#2668) 2016-09-13 11:01:10 -07:00
Kevin Qiu 241c56e6d7 Use Set instead of Add in the event that the request already contains the content-length (#2683) 2016-09-13 10:59:40 -07:00
Aditya Manthramurthy 895471afa1 Fixes #2678 (#2679)
Refactor `ratelimit.acquire()` to properly enforce the *globalMaxConn*
limit.
2016-09-12 15:53:54 -07:00
Harshavardhana 85e2d886bc bucketPolicy: Trim incoming paths properly for policy verification. 2016-09-11 10:36:54 -07:00
Harshavardhana 07aa02f196 build: Fix appveyor.yml to report windows coverage as well. (#2666) 2016-09-10 21:05:00 -07:00
Anis Elleuch 8acf4d112a Return Owner information in ListObjectV2 only when fetch-owner is specified (#2654) 2016-09-10 10:44:38 -07:00
Anis Elleuch e6abfb3b67 More tests for format-config for a better coverage (#2655) 2016-09-10 09:54:22 -07:00
Harshavardhana 040af08473 server: Startup message now prints configured ARNs. (#2653)
Fixes #2628
2016-09-10 02:23:28 -07:00
Harshavardhana e3de237eb8 Move config to v7 to fix previous v6 bug on null Notification entries. (#2650)
Thanks to @koolhead17 for reporting this.
2016-09-10 00:51:25 -07:00
Anis Elleuch 11785529fc Add Request Not Ready error when presigned date request is not valid (#2646) 2016-09-10 00:38:07 -07:00
Anis Elleuch 32201a18ab Forbid requests generated at least 15 min in the past or in the future (#2648) 2016-09-09 15:33:35 -07:00
Anis Elleuch 421cccb1d7 Add naughty disk to simulate disk errors (#2645) 2016-09-09 12:53:09 -07:00
Krishna Srinivas 400e9309f1 Signature-V4: Include content-length for signature calculation. (#2643)
This is to be compatible with clients which includes content-length for signature calculation
(and hence deviating from AWS Signature-v4 spec)
2016-09-09 09:38:07 -07:00
Krishna Srinivas a4afb312d4 S3-layer: Multiple-delete should return success for non-existent objects. (#2633)
Fixes #2630
2016-09-07 11:49:12 -07:00
Harshavardhana afe874f15a api: CopyObject pipe should be closed pro-actively upon error. (#2625)
Fixes a deadlock reproduced while running s3verify during
RemoveObject().  Previously held lock by GetObject() inside
the go-routine was never relenquished.
2016-09-06 02:23:32 -07:00
Harshavardhana b291dbe9c5 s3: Fix compliance bugs reported by s3verify. (#2624)
- Fixes couple of error strings reported are mismatching.
- Fixes a error HTTP status which was wrong fixed.
- Remove usage of an deprecated PostResponse, au contraire
  to their documentation there is no response body in
  PostPolicy.
2016-09-06 02:18:02 -07:00
Harshavardhana bc8f34bfe7 server/mux: Remove unused waitgroup from listenerMux. (#2614)
Change struct names to be more meaningful.
2016-09-02 19:07:42 -07:00
Harshavardhana 2dc7ecc59b notifications: Fix bucket notifications for DeleteMultipleObjects. (#2609)
Now reports bucket notifications for DeleteMultipleObjects API.
Also deletes multiple objects in parallel.
2016-09-02 01:59:08 -07:00
Hugo Jobling 600551feb9 Make export directory an exposed volume (#2610)
This exposes the /export directory, making it possible to mount it from other
containers. It also makes it possible to bind a local folder to the internal /export
mount point using the native Docker for Mac/Windows and Kinematic.
2016-09-02 01:58:15 -07:00
Harshavardhana 7398d737b5 profiler: Fix it properly and generate/save profiles even failure situations. (#2607)
Fixes #2594
2016-09-01 20:13:11 -07:00
Harshavardhana bf62ba57cf docs: Update minimum drives in erasure docs. 2016-09-01 11:35:53 -07:00
Harshavardhana 511a8cbe04 profiler: Fix how we generate profiles. (#2596)
Fixes #2594
2016-09-01 03:50:31 -07:00
koolhead17 790ad68d4d Doc: Fixed Minio install steps from source. (#2590) 2016-08-31 18:00:11 -07:00
Anis Elleuch e79d2381fc Fix rare 'go test -race' failure in ListenServe{Plain,TLS} (#2588) 2016-08-31 16:07:44 -07:00
Krishna Srinivas a3b4199e9b FS/Multipart: Allow for parallel uploads of parts with same part number. (#2587)
CBL client does not close connection when the backup process is stopped, this causes
read() on the stream on the server side to block and hence the lock held on the part
is not released. When the backup process is restarted, we again try to lock on the
part and this will block. Using a unique tmp name and not locking it fixes the problem.
2016-08-31 16:01:32 -07:00
Krishna Srinivas 3aa0574c6b FS/multipart: Append the parts to a file as and when the parts arrive. (#2513) 2016-08-31 13:42:57 -07:00
Andrew Stone 3118195e92 docker: Minimize image size (#2585)
This takes the image size from 823.1MB (current edge) to 281.8MB.

The following steps are taken:
  0) Use go-alpine for a much smaller base image
  1) Perform all actions in 1 step to eliminate diff images
  2) Remove unused files created after setup
2016-08-31 13:32:41 -07:00
Harshavardhana 5716f1c199 Fix incorrect error strings. (#2580) 2016-08-30 13:47:09 -07:00
Harshavardhana 2020afd200 server: http.Server do not add deadlines causes issues. (#2581)
Adding deadlines is a no go since Golang doesn't back off
the timers if there is an active i/o in progress.

It is meant to be for applications to handle this themselves
and manually progress the deadlines.

Fixes #2561
2016-08-30 12:52:19 -07:00
Bala FA 7431acb2c4 web: add handlers for set/get bucket policy. (#2486)
Refer #1858
2016-08-30 10:04:50 -07:00