[tahoe-dev] multi-user tahoe node suggestions

Zooko Wilcox-O'Hearn zooko at zooko.com
Wed Jan 13 18:45:21 UTC 2010

On Wednesday, 2010-01-13, at 1:11 , Kyle Markley wrote:

> What are the recommended methods for setting up a tahoe node to be  
> used by multiple users on a host, without using root?
> In my circumstance it's okay for a normal user (i.e., me) to run  
> the tahoe node with a crontab entry "@reboot tahoe start".

Go ahead and create a dedicated user account to run the tahoe-lafs  
web gateway process.  The web gateway doesn't need access to your  
personal files.

> How then do I best enable other users on the same machine to use  
> the same tahoe node?  (Is that even what I really want?)  In this  
> case I do want to enable convergence.

You let other people use your web gateway just by giving a URL to it  
-- http://$HOSTNAME_OR_IPADDR:$PORT where $PORT is determined by the  
"web.port" configuration: http://allmydata.org/trac/tahoe/browser/ 

You enable convergence with other people by making sure that the web  
gateway that they use and the one that you use has the same added  
convergence secret, found in a file named "private/convergence".  
configuration.txt#L408 .  If you and they are using hte same web  
gateway then you always get convergence with them.

> <p class="error"><class  
> 'allmydata.mutable.common.UncoordinatedWriteError'>: </p></ 
> a><div class="stackTrace"></div><a name="tracebackEnd"><p  
> class="error"><class  
> 'allmydata.mutable.common.UncoordinatedWriteError'>: </p>
> When I look at the node's status page and pull up the "Mutable File  
> Publish Status" page, it appears to have placed 3 of the shares.  I  
> can only guess about which file the error occurs on, but it is  
> reproducible.  I'm running allmydata-tahoe: 1.5.0 on the production  
> grid.

This is a combination of a bug in Tahoe-LAFS and degraded service of  
the allmydata.com prod grid.  Please help us track it down!   I've  
created a ticket to track it -- #899.  Please look in your ~/.tahoe/ 
logs/ directory and tar up the "incidents" subdirectory and attach  
the tarball to the ticket.  Also if you can paste in the "Mutable  
File Publsh Status" page and any other diagnostic information that  
you can get from the web user interface that would be appreciated.



http://allmydata.org/trac/tahoe/ticket/899#UncoordinatedWriteError on  
prod grid

More information about the tahoe-dev mailing list