Details
-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
None
-
1
Description
As a effort to make changes to system test for durability. We ran following test and observed that when we load data with persist-to 1 , then ops rate is very low 500 ops as compare to replicate-to ~50k ops.
Test with H/w- 4Node cluster (KV), 1 default bucket (1 replica)
Pillow fight
Standard(without persist to and replicate to) - ops - 90k-100k
Persist to 1 - Batch size 1000 (ops 500)
Replicate to 1 - 50k-75k
Gideon (https://github.com/couchbaselabs/gideon/blob/master/loader.py)
Standard(without persist to and replicate to)- ops - 50K
Persist to 1 - ~1500 ops
Replicate to 1 - 50K (edited)
these machines have SSDs(Samsung Pro 850) and 64 GB RAM, 48 vCPU
Docker commands for data loading
docker run sequoiatools/pillowfight:persist_replicate -U couchbase://172.23.104.16/default?select_bucket=true -B 1000 -I 1000000 -t 1 -P password --replicate-to 1
|
docker run sequoiatools/pillowfight:persist_replicate -U couchbase://172.23.104.16/default?select_bucket=true -B 1000 -I 1000000 -t 1 -P password --persist-to 1
|
docker run sequoiatools/pillowfight:persist_replicate -U couchbase://172.23.104.16/default?select_bucket=true -B 1000 -I 1000000 -t 1 -P password
|
docker run sequoiatools/gideon:persist_replicate kv --ops 10000 --create 100 --hosts 172.23.104.16 --bucket default --replicate_to 1
|
docker run sequoiatools/gideon:persist_replicate kv --ops 10000 --create 100 --hosts 172.23.104.16 --bucket default --persist_to 1
|
docker run sequoiatools/gideon:persist_replicate kv --ops 10000 --create 100 --hosts 172.23.104.16 --bucket default
|
