Less no of records being inserted in cassandra DB

[SourabhSanghi]

Hi,
I am trying to insert 1 Billion records to stress test Cassandra DB using multiple ycsb clients. I observe that less than 1 Billion records are inserted when i sum up the insert record count from each individual client. Is it a know issue with ycsb?? there is a loss of about 100 to 500 records overall.

[ghost]

Hi Sourabh,

I'm not sure what happened. Does it happen only when you use multiple ycsb clients?

--Michi

[SourabhSanghi]

Hi Michi,
I tried to isolate the issue and found out that it is not because of
multiple clients. Even when I run single ycsb client to load my Cassandra
cluster I am facing the same issue.

If for example i give a insertstart=0 and insertcount=58823529 it is
performing only 58823500 operations.

Regards,
Sourabh Sanghi

On Tue, Apr 24, 2012 at 1:47 AM, Michi Mutsuzaki <
reply@reply.github.com

wrote:

Hi Sourabh,

I'm not sure what happened. Does it happen only when you use multiple ycsb
clients?

--Michi

---

Reply to this email directly or view it on GitHub:
#74 (comment)

[SourabhSanghi]

and it's not a problem with just that particular record count 58823529 .
when i am trying with insertstart=0 insertcount=990 it is doing only 900
operations.

What i get from the experiments that i have done with YCSB is that it is
rounding off the insertcount to the with last 2 digits as 0's

Regards,
Sourabh Sanghi

On Tue, Apr 24, 2012 at 1:47 AM, Michi Mutsuzaki <
reply@reply.github.com

wrote:

Hi Sourabh,

I'm not sure what happened. Does it happen only when you use multiple ycsb
clients?

--Michi

---

Reply to this email directly or view it on GitHub:
#74 (comment)

[ghost]

Hi Sourabh,

Thanks for the info. It does sound like a bug in YCSB. Do you think you have time to create a patch for this?

Thanks!
--Michi

[kruthar]

@SourabhSanghi - sorry for the long pause on this issue. I tried to replicate with YCSB 0.5.0, Cassandra 2.1, and the cassandra2-cql client, but was unable to reproduce.

I tested with:

insertstart=0
insertcount=990

the record count seemed to be correct. Output excerpt from YCSB:

[INSERT], Operations, 990.0
[INSERT], AverageLatency(us), 2646.6656565656567
[INSERT], MinLatency(us), 1549.0
[INSERT], MaxLatency(us), 32399.0
[INSERT], 95thPercentileLatency(us), 3605.0
[INSERT], 99thPercentileLatency(us), 9327.0
[INSERT], Return=OK, 990

and cqlsh output:

cqlsh:ycsb> select count(*) from usertable;

 count
-------
   990

Is this still an issue?

[busbey]

closing as stale. please reopen if you're still seeing the issue with the changes that @kruthar noted

[ashleyblackmore]

I saw this in a very early version of YCSB, and dealt with it in typical student manner, by discarding in my application what were NPEs, iirc. I remember it was fixed in a later YCSB version.