Yehuda Sadeh
07b21866a1
rgw: update bucket instance with resharding details
...
can be useful info for later in a cross zone sync
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:59 -07:00
Yehuda Sadeh
40faa8d952
radosgw-admin: reshard execute output adjustments
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:59 -07:00
Yehuda Sadeh
aa3f71d6b1
cls/rgw: remove unused field in reshard entry
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:58 -07:00
Yehuda Sadeh
1da369e1e8
rgw: reshard: renew lease and handle marker when listing reshard repo
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:58 -07:00
Yehuda Sadeh
7054308e85
cls/rgw: fix reshard_list truncation identification
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:58 -07:00
Yehuda Sadeh
6cb30f9c95
cls/rgw: expose api to generate reshard indexing key
...
So that we could use it as a marker when doing reshard_list.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:57 -07:00
Yehuda Sadeh
81693efd78
rgw: remove unneeded and duplicate code
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:57 -07:00
Yehuda Sadeh
6c540bd5fe
rgw: initialize bucket structure
...
after reading bucket info
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:56 -07:00
Yehuda Sadeh
485a11fe59
rgw-admin: revive reshard execute command
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:56 -07:00
Yehuda Sadeh
ed72675ba2
rgw: bucket_id does not include bucket name
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:55 -07:00
Yehuda Sadeh
b516c92af5
rgw: bi_list() initialize is_truncated if -ENOENT
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:54 -07:00
Yehuda Sadeh
999cca4ed3
rgw: internal api cleanup
...
rename method, minor change
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:54 -07:00
Yehuda Sadeh
436f151e55
cls/rgw: reshard add command doesn't read existing status
...
No need to read existing status.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:54 -07:00
Yehuda Sadeh
2da8dc5d10
cls/rgw: fix remove_reshard objclass op
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:53 -07:00
Yehuda Sadeh
f677d0e960
cls/rgw: modify logshard key
...
The logshard key shouldn't include bucket id.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:53 -07:00
Yehuda Sadeh
8e1bf1b9e8
rgw: multiple fixes and adjustments related to resharding scheduler
...
still wip, but:
- get rid of some unneeded rados locking
- rename shards to logshards where relevant to avoid confusion
- remove unused code
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:53 -07:00
Yehuda Sadeh
b757e7e16e
rgw: fix compilation following rebase
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:52 -07:00
Yehuda Sadeh
cac047a88b
rgw: replace reshard blocking sleep with interruptible condition wait
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:52 -07:00
Orit Wasserman
7114124673
rgw: actually reshard an entry from the resharding queue
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:52 -07:00
Orit Wasserman
c7fc87f9f1
rgw: check if buckets needs resharding on all ops that add new objects
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:51 -07:00
Orit Wasserman
444b9c918c
rgw: add bucket to resharding queue if needs resharding
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:51 -07:00
Orit Wasserman
e03b85e987
rgw: check for bucket resharding when adding an object
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:51 -07:00
Orit Wasserman
9a084497fd
rgw: add a function to add a bucket to the resharding queue
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:50 -07:00
Orit Wasserman
553bf4b996
rgw: add a check_bucket_shards to the quota
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:50 -07:00
Orit Wasserman
1ed76a680c
rgw: moved reshard_bucket into RGWReshard
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:49 -07:00
Orit Wasserman
dfdc2b2cc0
rgw: remove cct from RGWReshard and BucketIndexLockGuard
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:49 -07:00
Orit Wasserman
ad138a4d72
rgw: add resharding thread
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:49 -07:00
Yehuda Sadeh
806d47329e
rgw: store new num shards in bucket index reshard entry
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:48 -07:00
Yehuda Sadeh
42b3ee12ea
rgw: fix deletion of index completion object
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:48 -07:00
Yehuda Sadeh
94e3c380ab
rgw: reshard complete handling, use multiple locks
...
to reduce contention. Also, fix completions cleanup.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:47 -07:00
Yehuda Sadeh
da1f5bc910
rgw: clean up index completions on shutdown
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:47 -07:00
Yehuda Sadeh
60778b0214
rgw: update data log after index completion reshard handling
...
Also rearrange code a bit.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:47 -07:00
Yehuda Sadeh
245cd0f9af
rgw: guard resharding in async completions
...
Catch return value from async index completions, resend operations
synchronously on a separate worker thread if needed.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:46 -07:00
Yehuda Sadeh
e18e271490
rgw: reshard guard for versioned operations
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:46 -07:00
Yehuda Sadeh
9677f7d31c
cls/rgw: changes to support reshard guard on olh bi ops
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:46 -07:00
Yehuda Sadeh
8fd589d1eb
rgw: handle reshard wait logic in cancel() too
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:45 -07:00
Yehuda Sadeh
d250fab244
rgw: fix update_bucket_id, reshard blocking logic
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:45 -07:00
Yehuda Sadeh
9766f5b92d
rgw_admin: add rgw-admin reshard status command and initial blocking logic
...
A command that returns the reshard status field on the bucket index objects (all
shards).
Also, initial implementation of identifying and blocking while reshard in
progress.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:44 -07:00
Yehuda Sadeh
09ccfebf68
rgw: more detailed resharding status kept in bucket header
...
So that we can differentiate between in-progress and complete status.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:44 -07:00
Yehuda Sadeh
e4d4aa8b86
rgw: a few fixes
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:44 -07:00
Yehuda Sadeh
660a512635
rgw: mark source bucket shards when starting reshard
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:43 -07:00
Yehuda Sadeh
bb0cb4d7a5
rgw: move resharding code to BucketReshard class
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:43 -07:00
Yehuda Sadeh
7a34a049ae
rgw: split per-bucket resharding logic out from general resharding
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:43 -07:00
Yehuda Sadeh
0bc15671ee
rgw: revert some reshard blocking logic
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:42 -07:00
Yehuda Sadeh
7d44601e83
cls/rgw: create an objclass method to guard in-progress resharding
...
This method will be called on the bucket index shard.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:42 -07:00
Orit Wasserman
4ddb785bb0
rgw: RGWReshard update to new cls API
...
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
2017-06-05 13:17:41 -07:00
Yehuda Sadeh
a07e8eb6af
cls/rgw: adjust types and ops
...
- don't keep data bufferlist in struct cls_rgw_bucket_instance_entry
- remove rgw_get_bucket_instance
We need to keep the reshard info as part of the bucket index header (for
each source bucket shard), and not in a separate object. Otherwise it'll
be impossible to have the resharding check atomic as part of the regular
bucket index updates.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:41 -07:00
Yehuda Sadeh
04cf0ca074
cls/rgw: minor changes
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:41 -07:00
Yehuda Sadeh
44e46b35f0
cls/rgw: generate_reshard_key, change output param
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:40 -07:00
Yehuda Sadeh
f446b80e2a
cls/rgw: return default header if not initialized
...
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
2017-06-05 13:17:40 -07:00