Re: Select with 5K+ columns kills server

On 01 Oct 2016, at 05:54, Yves Dorfsman wrote:

On 2016-09-30 16:36, Pascal Heus wrote:

...

Now this is on a AWS EC2 t2.small instance with just ~2Gb or memory. Maybe more memory may solve this issue (?). But in any case concerned that the server hangs without warning or error message in the log.

You mean the merovingian.log? I’d expect mserver5 to give errors like HEAPextend failed. But I doubt if that error message goed into the merovingian.log Probably, mserver5 doesn’t really hang, but it can’t do anything. Since the error message of mserver5 probably doesn’t show up in the merovingian log, it feels like that the server hangs. You can wait for some time (30+ sec) for the server to clean up its temporary data, and try a ‘select 1’ query to check if the server is really not responding… Jennie

...

Any potential workaround? Suggestion most appreciated

How many rows are you talking about?

t2s are very limited on CPU and I/O bandwidth, they also reduce your CPU power once you've reached your limit. This also comes faster than you'd expect, we used to run into this on small web server from compressing the log files.

I find bigger instances (m3s) better for monetdb, especially that being analytic work, we can shut down the instances when not in use, keep the cost to something very reasonable.

-- http://yves.zioup.com gpg: 4096R/32B0F416

_______________________________________________ users-list mailing list users-list@monetdb.org https://www.monetdb.org/mailman/listinfo/users-list