Impact of queries while redistribution is active
Article ID: KB0084341
Updated On:
| Products | Versions |
|---|---|
| TIBCO ActiveSpaces | 2.2.1 |
Description
Redistribution takes very long to finish. A non-indexed query for large Spaces (high CPU usage) seems to block all Space activties.
Issue/Introduction
Redistribution seems to hang when perform a non-indexed query for large Spaces.
Environment
All Operating Systems
Resolution
Redistribution starts when Seeders leave or join to the metaspace.
When a query will be executed at this time, redistribution stops until the query finishes.
At this time the as-admin command "show space '<spacename>' stats" shows no write operations (put operations are slow/hang).
On large Spaces, a non-index query can block all operations caused by a full table scan (pending query increasing).
The redistribution and query actually use the same thread, which means that redistribution will wait until the query finishes.
To avoid slowness, be sure an index query is used on large metaspaces.
When a query will be executed at this time, redistribution stops until the query finishes.
At this time the as-admin command "show space '<spacename>' stats" shows no write operations (put operations are slow/hang).
On large Spaces, a non-index query can block all operations caused by a full table scan (pending query increasing).
The redistribution and query actually use the same thread, which means that redistribution will wait until the query finishes.
To avoid slowness, be sure an index query is used on large metaspaces.
Feedback
Yes
No
Powered by
