[chord] replication in chord + dhash

Srinivas V. srini at cs.arizona.edu
Tue May 27 09:22:19 EDT 2003


Hello


I recently built and set up CFS and have been trying to get it up and
running. I have been running the lsd daemon on a few machines and
storing/fetching blocks using the dbm client.

Initially I ran some simple tests where I started lsd on two hosts. I used
the dbm program on the hosts to store and fetch blocks i.e. I would store
blocks on the ring in one host and was able to fetch them using dbm on the
other host.

When I kill lsd, on one of the the hosts and use dbm on the other host to
try fetching blocks, some of the blocks are unavailable.I get error
messages like:

dbm: dhashclient::retrieve failed: 24ec132e88b234835fd9aa3f7c9c4fa91357d3f4: DHASH_NOENT
dbm: dhashclient::retrieve failed: 7b94623bccedaf84c537684d62a21b9ff51d2b99: DHASH_CHORDERR

I noticed that in lsd, one could pass the replication factor with the -n
command line flag (I didn't set this initially since lsd's usage message
didn't display this option). In later tests I did use this flag to set the
replication factor at 8. For each of the two chord nodes I started, I set
the no. of virtual nodes to 16. However, I still get the error message
about blocks being unavailable. while starting lsd, I use the following
options/flags:
./lsd -d /tmp/db-something -j host:port [-p port_for_bootstrap] -v vnodes -n nreplicas

I wanted to know if I was doing anything wrong and if there's anything
else I have to set.

     thanks

     srinivas




More information about the chord mailing list