Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.5.0
-
CB build: 5.5.0-1607
OS: CentOS 7
CPU: E5-2680 v3 (48 vCPU)
Memory: 64GB
Disk: Samsung PM863a
-
Untriaged
-
Centos 64-bit
-
No
-
CBSS Sprint Feb-16-2018
Description
Sample doc:
{
"_sync": {
"rev": "1-6818bd58742046269be08ffdf0268bb2",
"role_access": {
"sg-user-6": {
"sg-role-0": 142613,
"sg-role-1": 142613,
"sg-role-7": 142613,
"sg-role-9": 142613
…
The are 1M such documents in the bucket. The "role_access" field has 1000 elements average.
Index definition:
CREATE INDEX `ixRoleAccess` ON `bucket-1`((distinct (array (`op`.`name`) for `op` in object_pairs(_sync.`role_access`) end)),object_pairs(_sync.`role_access`));
Index node has 64 GB RAM total.
35 GB RAM has been allocated to the index.
Once I run create index query the indexer quickly uses all 64 GB RAM, runs out of memory and indexer restarts.
On the time the crash happens the index actual size is only 5GB and its about 40% done.
When using MOI everything works as expected. Total size of the index when completed is about 20 GB.
It works fine also with lesser arrays (1, 10, 100 elements)
Attachments
Issue Links
- mentioned in
-
Page Loading...
For Gerrit Dashboard: MB-27425 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
90639,2 | MB-27425: Add a simple script to load data into couchbase | master | gocb | Status: ABANDONED | 0 | 0 |