Forum: CFEngine Help
Subject: Re: Cfengine 3.3 wins!
Author: msvob...@linkedin.com
Link to topic: https://cfengine.com/forum/read.php?3,25626,26049#msg-26049
I think it was the move from the Berkeley DB backend database to Tokyocabinet.
$ /cfengine/trunk
msvoboda-4-mn:trunk msvoboda$ find . | wc -l
7902
So I have nowhere near that many objects. I remember talking to Diego and
Eynstein saying while they were visiting LinkedIn that large amounts of small
files is a lot more work for cf-serverd than a few large ones. A MD5 sum has
to be computed for each file / exchange between cf-serverd and cf-agent.
An alternative you might have is to automate some sort of "tarball" process for
all of your small data if possible. We do this to automate some of our tool
releases that we push to all infrastructure and dont want Cfengine to have to
deal with performing static digest management of every single file. An
automated process that generates a tarball also generates a file containing a
MD5 sum. I have cf-agent compare the MD5 sum it reads from that file against
the actual MD5 sum value of the tarball. If they match, I extract and deploy
via cf-agent.
It might be a way to help offload so much work being done by cf-serverd, yet
still guarantee that your data is intact via MD5 checksumming. You'd only have
to perform a MD5 checksum over 1 file instead of 200k.
_______________________________________________
Help-cfengine mailing list
Help-cfengine@cfengine.org
https://cfengine.org/mailman/listinfo/help-cfengine
