Hi Sean, thanks for the very quick response :) - much appreciated. No, I didn't put any retries in the code at all.
So, if I write a key to a cluster and then try to query that key, it may not find it unless I retry? Even if r=2 & w=2 ? I thought that at least 2 systems had to have that key (either in memory or disk) for the write to have succeeded with those settings? If I had w=3, would you expect any missed keys ?? (note that I didn't try that because I expected performance issues). Is it possible that I'd need to retry more than once? I guess the issue I have is (when it comes to real world usage) when do I give up checking if a key is there - it might genuinely not be there, or just not replicated yet? I'll be running the tests again after the w/e (probably Monday or Tuesday), so hopefully I'll get more information there (and add a retry :)) thanks again, Pat. > Date: Fri, 14 Sep 2012 10:50:06 -0400 > Subject: Re: missing keys? > From: [email protected] > To: [email protected] > CC: [email protected] > > Pat, > > Does your test use retries when it gets "not found" responses? You > might not encounter that small % of missing keys if you retry at least > once when they are missing. Because of the eventually consistent > design of Riak, writes you make might not be readable immediately. > > On Fri, Sep 14, 2012 at 10:17 AM, Pat Lynch <[email protected]> wrote: > > Hi all, > > I was just wondering if you could give me some feedback on my usage of RIAK. > > I'm currently testing a number of datastores and am trying to come up with > > pros/ cons of each. > > > > I have a variety of tests (write speed 1 client, write speed 10 clients, > > read speed 1 client, read speed 10 clients, read while writing, write while > > reading etc). The data set is tiny for this phase (10k, 50k, 100k). > > > > I have 3 machines in the cluster (each with 12GB RAM, quadcore CPUs, 500GB > > drives), each cache setup to use 128MB memory (eleveldb). Each key is 64 > > bytes. and each value is 512 bytes. > > > > > > The write speed I'm getting is around 260.0440791 per second (using 1 > > client), scaling to 2281.1662 writes per second (using 10 clients). > > The read speed I'm getting is around 397.2291042 per second (using 1 > > client), scaling to 3316.089445 reads per second (using 10 clients). > > > > One very worrying thing I noticed is that I sometimes miss keys? i.e. I > > write 100k keys to disk, then read them all back in and check the value > > matches expected. This would only happen for 1 or 2 keys out of 100k, but > > (1 or 2 keys is enough…) > > > > This is using the latest version of RIAK with the python client (python > > client is running on a 4th machine). > > > > Any help would be greatly appreciated. > > > > thanks, > > Pat. > > > > _______________________________________________ > > riak-users mailing list > > [email protected] > > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > > > > > > -- > Sean Cribbs <[email protected]> > Software Engineer > Basho Technologies, Inc. > http://basho.com/
_______________________________________________ riak-users mailing list [email protected] http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
