Details
-
Bug
-
Resolution: Unresolved
-
Critical
-
6.6.3, 7.0.2, 7.1.0
-
Untriaged
-
1
-
No
Description
Steps to repro:
a. Create a 1 node KV + n1ql + index cluster
b. Load travel sample
c. Drop travel sample
d. Query the localhost:9102/getLocalIndexMetada endpoint
The output contains topologies as under:
"topologies" : [
|
{
|
"bucket" : "travel-sample",
|
"version" : 8,
|
"scope" : "inventory",
|
"collection" : "airline"
|
},
|
{
|
"collection" : "_default",
|
"scope" : "_default",
|
"version" : 70,
|
"bucket" : "travel-sample"
|
},
|
This is a leak as the topology information is never created and the meta size grows. Also, in monitorKeyspace(), we read repo by topology. Since this topology exists in repo but it is actually deleted, we go and query ns_server thinking our cache is stale. ns_server says it is removed and we do nothing. 10 seconds later, same process happens - flooding ns_server with un-necessary requests
Attachments
Issue Links
- duplicates
-
MB-50831 Avoid call to ns_server when querying info for deleted bucket
- Closed
Gerrit Reviews
For Gerrit Dashboard: MB-50790 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
198529,4 | MB-50790 Delete topology from local repo when keyspace is deleted | unstable | indexing | Status: NEW | 0 | 0 |