Hi Ken.
1) Thanks for the great link.
Ironically it was written by Netflix, who continued to use single tokenfor
years after vnodes were released so that they could continue touse Priam and
their other tools dependent on single token. (I was in theearly Cassandra group
there.)
2) My tool agrees overall with their findings:
a) it does reflect that increasing numbers of vnodes and nodes reduce
reliabilitydramatically, so the results are conceptually the same and the
deltas atdifferent vnode counts matches what I see in my calculator.
b) but it uses a more complicated model. I'm happy with my calculator thatlooks
at simple "probability of a streaming connection failed for any reason"and is
immediately usable by any DBA or SRE.
3) As an Operations DBA, their reference to "centuries" made me laugh
though.Note that my calculations are about failures within one week, which
alignsmore with my experience. So either they're overly optimistic, or I'm
pessimistic.
You can verify which by doing a grep of your logs on a production cluster fora
month and counting how many connection failures there were. My blogpost has
some links to actual error message to grep for. 4) Note that Datastax
recommends 8 vnodes now. See my blog for the reference.
Thanks, James Briggs.
--
Cassandra/MySQL DBA. Available in Bay area or remote.
cass_top: https://github.com/jamesbriggs/cassandra-top
From: Kenneth Brotman
To: user@cassandra.apache.org
Sent: Saturday, February 16, 2019 5:00 AM
Subject: RE: Cassandra vnodes Streaming Reliability Calculator
#yiv4674113709 #yiv4674113709 -- _filtered #yiv4674113709
{font-family:Calibri;panose-1:2 15 5 2 2 2 4 3 2 4;} _filtered #yiv4674113709
{font-family:Tahoma;panose-1:2 11 6 4 3 5 4 4 2 4;}#yiv4674113709
#yiv4674113709 p.yiv4674113709MsoNormal, #yiv4674113709
li.yiv4674113709MsoNormal, #yiv4674113709 div.yiv4674113709MsoNormal
{margin:0in;margin-bottom:.0001pt;font-size:12.0pt;font-family:New;}#yiv4674113709
a:link, #yiv4674113709 span.yiv4674113709MsoHyperlink
{color:blue;text-decoration:underline;}#yiv4674113709 a:visited, #yiv4674113709
span.yiv4674113709MsoHyperlinkFollowed
{color:purple;text-decoration:underline;}#yiv4674113709
span.yiv4674113709EmailStyle17 {color:#1F497D;}#yiv4674113709
.yiv4674113709MsoChpDefault {font-size:10.0pt;} _filtered #yiv4674113709
{margin:1.0in 1.0in 1.0in 1.0in;}#yiv4674113709 div.yiv4674113709WordSection1
{}#yiv4674113709 Hi James, Thanks for doing that. Very interesting. I
haven’t had a chance to check the math. Did you look at this white paper by
Lynch and Snyder called Cassandra Availability with Virtual Nodes:
https://github.com/jolynch/python_performance_toolkit/blob/master/notebooks/cassandra_availability/whitepaper/cassandra-availability-virtual.pdf
Are the calculations consistent with your online calculator? Thanks again,
Kenneth Brotman From: James Briggs [mailto:james.bri...@yahoo.com.INVALID]
Sent: Friday, February 15, 2019 7:42 PM
To: user@cassandra.apache.org
Subject: Cassandra vnodes Streaming Reliability Calculator Hi folks.
Please check out my online vnodes reliability calculator and reply with any
feedback:http://www.jebriggs.com/blog/2019/02/cassandra-vnodes-reliability-calculator/
Thanks, James Briggs.
--
Cassandra/MySQL DBA. Available in Bay Area or remote.
cass_top: https://github.com/jamesbriggs/cassandra-top