People tend to make configuration changes by zookeeper first, and then migrate them to the config files for CM/documentation purposes.
We could get the config files from HDFS, but that's not supported right now. Accumulo 1.6 can get jar files from there, for the reason you state: it simplifies deploying the correct file to all nodes in the cluster. -Eric On Fri, Jul 18, 2014 at 11:02 AM, Geoffry Roberts <[email protected]> wrote: > I cited the slaves file as no more than a possibility, hence the e.g. > > I remember, when I ran a Hadoop cluster, using rsync to deploy config > updates to all nodes. I assumed I would be doing the same with Accumulo > once I got into multi nodes. Apparently, from what you're telling, me > this copying remains the preferred practice despite what ZK offers. > Interesting. > > > On Fri, Jul 18, 2014 at 9:32 AM, Eric Newton <[email protected]> > wrote: > >> I don't know of anyone putting the slaves file in zookeeper. However, >> nearly all the basic settings can be configured in zookeeper: changing any >> setting in the shell puts the configuration change in zookeeper. >> >> That said, most people like to get their changes in the conf directory >> because those files can be managed easily with CM tools and install tools. >> >> -Eric >> >> >> >> On Fri, Jul 18, 2014 at 8:15 AM, Geoffry Roberts <[email protected]> >> wrote: >> >>> All, >>> >>> Is anybody using ZK to configure Accumulo? >>> >>> Heretofore, I've been focused on Accumulo and Zookeeper was just a means >>> to that end. Lately, I dug a bit into ZK and learned one of it's purposes >>> is to provide a common point of configuration. But in the Accumulo >>> documentation, I've seen nothing about using ZK in that way. Is anybody >>> using ZK to configure Accumulo? e.g. speccing the slaves list in ZK as >>> opposed to the conf directory. >>> >>> Thanks >>> >>> -- >>> There are ways and there are ways, >>> >>> Geoffry Roberts >>> >> >> > > > -- > There are ways and there are ways, > > Geoffry Roberts >
