See also #32440
- Add loadTask in collection observer
- For load collection/partitions, load task shall timeout as a whole
- Change related constructor to load jobs
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
See also #31632
Release conc.Pool after usage to clean worker and stop background purge
and ticktock.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #30647
- Same dst and src resource group should not be allowed in
`TransferReplica` and `TransferNode`.
- Remove redundant parameter check.
Signed-off-by: chyezh <chyezh@outlook.com>
Related to https://github.com/milvus-io/milvus/issues/32165
Avoid iterating through all replicas/collections if possible. Iteration
is expensive when there are large number of replicas/collections.
Signed-off-by: yiwangdr <yiwangdr@gmail.com>
Related to #32165
Add collection to partitionIDs mapping to avoid interation on all
partitions loaded when trying to get all partitions with collection id
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #30647
- Add declarative resource group api
- Add config for resource group management
- Resource group recovery enhancement
---------
Signed-off-by: chyezh <chyezh@outlook.com>
See also #32066
This PR make coordinator register successful and let
`ProcessActiveStandBy` run async. And roles may receive stop signal and
notify servers.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #31091
This PR add GetByFilter interface in leader view manager, instead of all
kind of get func
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30816
check stale rules for leader task:
1. for reduce leader task, it should keep executing until leader's node
become offline.
2. for grow leader task,it should keep executing until leader's node
become stopping.
This PR check leader node's stopping state for grow leader task
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #29507
Notice that api_testonly.go files should be guarded by compiler tag
`test`, so that production build rules don't compile them and these APIs
don't get misused.
Signed-off-by: yiwangdr <yiwangdr@gmail.com>
issue: #30647
- ReplicaManager manage read only node now, and always do persistent of
node distribution of replica.
- All segment/channel checker using ReplicaManager to get read-only node
or read-write node, but not ResourceManager.
- ReplicaManager promise that only apply unique querynode to one replica
in same collection now (replicas in same collection never hold same
querynode at same time).
- ReplicaManager promise that fairly node count assignment policy if
multi replicas of collection is assigned to one resource group.
- Move some parameters check into ReplicaManager to avoid data race.
- Allow transfer replica to resource group that already load replica of
same collection
- Allow transfer node between resource groups that load replica of same
collection
---------
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #31091
This PR add GetByFilter interface in channel dist manager, instead of
all kind of get func
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #31468
1. when segment's version in leader view doesn't match segment's version
in dist, should update leader view
2. after call loadDeltalog, should update segment's load version with
latest ts
3. change leader task's priority from high to low, to avoid leader task
replace segment task and balance task
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #29261
This PR Add restful api for devops to execute rolling upgrade, including
suspend/resume balance and manual transfer segments/channels.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #31480#31481
release duplicate l0 segment task, which execute on old delegator may
cause segment lack, and execute on new delegator may break new
delegator's leader view.
This PR skip release duplicate l0 segment by segment_checker, cause l0
segment will be released with unsub channel
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #28491#31240
When colleciton number is large, querycoord saves collection target one
by one, which is slow and may block querycoord exits.
In local run, 500 collections scenario may lead to about 40 seconds
saving collection targets.
This PR changes the `SaveCollectionTarget` interface into batch one and
organizes the collection in 16 per bundle batches to accelerate this
procedure.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #30816
pr #31319 introduce the logic that segment checker need to load level
zero segment which only exist in current target.
This PR fix load segment task promote failed when segment only belongs
to current target
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
This PR add metrics for task latency in querycoord scheduler, so if any
kind of task stuck, it's easy to figure out by metrics
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
"-1" as `InvalidPartitionID` previously used as All partition place
holder in delete cases. It's confusing and hard to maintain when a const
var has more than one meaning.
This PR add `AllPartitionsID` to replace these usages in delete
scenarios.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #31162
when give scope CurrentTargetFirst/NextTargetFirst, it's expected to
scan both current and next target.
This PR fixed wrong behavior of CurrentTargetFirst/NextTargetFirst in
target manager, which may cause unexpected task generated, and load
collection may stuck forever due to dirty leader view.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30186
during channel balance, after new delegator loaded, instead of syncing
l0 segment's location to new delegator, we should load l0 segment on new
delegator, and release the old l0 segment, then start to release old
delegator.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #28491
after querycoord restart, it will pull a new target, which include
channel and segment list. when segments loaded on querynode has reached
the target, the collection could provide search/query. but if segment
list changes by time, ater querycoord pull a new target, it will takes a
few minutes to catch up the target's segment distribution. and before
that, query/search will fail due to lack of segments.
This PR save the current loaded target to meta storein querycoord's stop
progress, and recover it when query coord starts, to speed up the target
recovery time.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30816
cause balance channel will stuck until leader view catch up the current
target, then start to unsub the old delegator. which make sure that the
new delegator can provide search before release old delegator. but
another logic in segment_checker skip loading segment during balance
channel. so during balance channel, if query node crash, new delegator
can't catch up target forever, then stuck forever.
This PR remove the rule that skip loading segment during balance channel
to avoid the logic dead lock here.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30983#30982
cause balancer call wrong interface to get segment/channel list in
replica, then got a wrong average segment/channel number, which make
each node have less segment/channel than average, and the balance won't
be trigger in multi replica case.
This PR fix that balance segment/channel won't be trigger on multi
replicas
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30890
when leader checker find that leader view has an older load version of
segment, it will try to correct leader view. but the sync action doesn't
specify the latest load version. so the update operation will failed.
This PR fix leader checker can't update segment's load version and
keeping generate same task to scheduler.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #31103
Since querycoord need index meta information from datacoord only, broker
shall use `ListIndexes` to skip segment index building check logic in
datacoord
This PR is also related to #30538, in which DescribeIndex caused lots of
memory usage and lead to OOM eventually
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
1. add coordinator graceful stop timeout to 5s
2. change the order of datacoord component while stop
3. change querynode grace stop timeout to 900s, and we should
potentially change this to 600s when graceful stop is smooth
issue: #30310
also see pr: #30306
---------
Signed-off-by: chyezh <chyezh@outlook.com>
issue: #30150
`checkLeaderTaskStale` will check segment whether exist on next current
for leaderTask's growing action, which will cause promote leader task
failed when segment only exist on current target
This PR will check segment for both current or next target.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30150
see also: #30258
cause `SyncDataDistribution` will try to load delta for segment. if miss
indexInfo in request, sync action will failed due to lack of index info.
This PR set indexinfo when try to set segment to leader view
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #30723
This PR skip generate balance task when collection's target isn't ready.
also refine the check stale logic in query coord's scheduler, if channel
exist in current or next target, task won't be canceled.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #30150
For leader view distribution with offline nodes, a release task can
never be sent to querynode due to targetNode online check logic. Even
the request is dispatched, normal release task does not have "force"
flag when calling `delegator.ReleaseSegment`.
This PR adds a new type of querycoord task: LeaderTask, the
responsibility of which is to rectify leader view distribtion.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #30553
when datacoord with version 2.2 and querycoord with version 2.3 coexist
during rolling upgrade, `DescribeIndex/GetIndexInfo` will return
`unimplemented` error
This PR add retry on `DescribeIndex/GetIndexInfo`, to prevent load
collection failed during rolling upgrade from milvus 2.2 to 2.3.
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
This PR mainly improve two items:
1. Target observer should refresh loading status during init time. An
uninitialized loading status blocks search/query. Currently, the target
observer refreshes every 10 seconds, i.e. we'd need to wait for 10s for
no reason. That's also the reason why we constantly see false log
"collection unloaded" upon mixcoord restarts.
2. Delete session when service is stopped. So that the new service
doesn't need to wait for the previous session to expire (~10s).
Item 1 is the major improvement of this PR, which should speed up init
time by 10s.
Item 2 is not a big concern in most cases as coordinators usually shut
down after stop(). In those cases, coordinator restart triggers serverID
change which further triggers an existing logic that deletes expired
session. This PR only fixes rare cases where serverID doesn't change.
integration test:
`go test -tags dynamic -v -coverprofile=profile.out -covermode=atomic
tests/integration/coordrecovery/coord_recovery_test.go -timeout=20m`
Performance after the change:
Average init time of coordinators: 10s
Hardware: M2 Pro
Test setup: 1000 collections with 1000 rows (dim=128) per collection.
issue: #29409
Signed-off-by: yiwangdr <yiwangdr@gmail.com>
issue: #30150
This PR fix three problems:
1. leader checker use wrong node id when generate release task, which
cause the release task finished immediately
2. the release request generated by leader_checker doesn't set the
`force` flag, the operation to clean leader view on delegator will fail.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
don't store logPath in meta to reduce memory, when service get
segmentinfo, generate logpath from logid.
#28885
Signed-off-by: lixinguo <xinguo.li@zilliz.com>
Co-authored-by: lixinguo <xinguo.li@zilliz.com>
issue: #29841
This PR fix leader checker use wrong check interval, which causes leader
checker trigger too frequency
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #29841
if segment loaded, submit load segment task for it isn't permitted, to
avoid load segment twice. but this logic blocks the leader checker to
correct leader view by `LoadSegment`
This PR remove the segment loaded check, to fix that leader checker
cann't submit load task
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
issue: #29677#29838
during get shard leaders, if qeurynode doesn't ack the heartbeat than
10s, querycoord will treat it as unavailable, and won't return shard
leader on it. but when querynode has a full cpu usage, it's easily to
stuck for more than 10s without ack the heartbeat, which cause no shard
leader to search/query.
This PR remove heartbeat lag logic during get shard leaders
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
the recent changes move the level 0 segments list to a new proto field,
which leads to the QueryCoord can't see the level 0 segments, handle the
new changes
fix#29907
Signed-off-by: yah01 <yang.cen@zilliz.com>
See also: #29650
Either segment dml position & channel checkpoint could be newer in some
cases. This PR make PackLoadSegments use the newer one improving load
performance during cases where there are lots of upsert.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #29814
if channel is not subscribed yet, the generated load segment task will
be remove from task scheduler due to the load segment task need to be
transfer to worker node by shard leader.
This PR skip generate load segment task when channel is not subscribed
yet.
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #29803
This PR:
- Add trace span for collection/partition load
- Use TraceSpan to generate Segment/ChannelTasks when loading
- Refine BaseTask trace tag usage
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #29453
sync distribution by rpc will also call loadSegment/releaseSegment,
which may cause all kinds of concurrent case on same segment, such as
concurrent load and release on one segment.
This PR add leader_checker which generate load/release task to correct
the leader view, instead of calling sync distribution by rpc
---------
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
See also #29699
Querycoord panicked when tried to pop from an empty heap. We assume the
heap shall not be empty, but in some branch, the candidate is never
pushed back.
This PR put pop & push in a closure and adds a defer call to push item
back.
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
See also #29625
This PR:
- Add a new implemention of `DeleteBuffer`: listDeleteBuffer
- holds cacheBlock slice
- `Put` method append new delete data into last block
- when a block is full, append a new block into the list
- Add `TryDiscard` method for `DeleteBuffer` interface
- For doubleCacheBuffer, do nothing
- For listDeleteBuffer, try to evict "old" blocks, which are blocks
before the first block whose start ts is behind provided ts
- Add checkpoint field for `UpdateVersion` sync action, which shall be
used to discard old cache delete block
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
See also #27675
For `GetRecoveryInfo` & `GetRecoveryInfoV2`, Level zero segment ids
shall be specified in vchan info so that querycoord could re-fetch
current segment info during watch procedure without having all segment
info
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
Related to #29575
Add `getCollectionTarget` method which is atomic when scope is
`CurrentTargetFirst` or `NextTargetFirst`
Also return error when executor finds no channel in target manager
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
issue: #29523
readable shard leader should still be the old one during channel
balance, if the new shard leader is not ready.
This PR fixed that query coord choose wrong shard leader during balance
channel
Signed-off-by: Wei Liu <wei.liu@zilliz.com>
`WatchDmChannel` only need growing segment info, this PR removes fetch
segmentInfos when fill watch dml channel request.
---------
Signed-off-by: Congqi Xia <congqi.xia@zilliz.com>
