Also it is probably worth to mention:
1. I see no other errors in logs except that one;
2. Sometimes connected clients receive "Request did not complete within
rpc_timeout.", even if they are accessing other tables.
3. Sometimes, some cells from another tables may read as NULL when they are
in fact
Cassandra 1.2.8 still have this issue.
Possible recipe to reproduce: create the table as described in the first
message of this thread; write 3000 rows of 10MB each at the rate about
0.1..1 request per second.
Maybe this behavior is caused by incremental compaction of large rows...
On Mon, Jul 2
Great. Let me know what you find!
Thanks!
Paul
Sent from my iPhone
On Jul 27, 2013, at 2:47 AM, Pavel Kirienko
wrote:
> Hi Paul,
>
> I checked out your issue, looks the same indeed. Probably this can be reproed
> simply by writing large rows (> 10MB) on high rates.
> I'm going to try 1.2.7
Hi Paul,
I checked out your issue, looks the same indeed. Probably this can be
reproed simply by writing large rows (> 10MB) on high rates.
I'm going to try 1.2.7 then will be back with results.
On Sat, Jul 27, 2013 at 12:18 AM, Paul Ingalls wrote:
> This is the same issue we have been seeing.
This is the same issue we have been seeing. Still no luck getting a simple
repro case for creating a JIRA issue. Do you have something simple enough to
drop in a JIRA report?
Paul
On Jul 26, 2013, at 8:06 AM, Pavel Kirienko
wrote:
> Hi list,
>
> We run Cassandra 1.2 on three-node cluster.
Hi list,
We run Cassandra 1.2 on three-node cluster. Each node has 16GB RAM, single
200GB HDD with Ubuntu Server 12.04.
There is an issue with one table that contains about 3000 rows, here its
describe-table:
CREATE TABLE outputs (
appid text,
staged boolean,
field ascii,
data blob,
PR
