0

There are some wiki which explain how to install and setup AppScale.

But I didn't find anything related to the extensive use of AppScale. I mean, what to do in case of major outage ?

  • If the node crashes but can be rebooted and could work : does AppScale automatically replicates what is missing from other nodes to it ? So basically, as long as the node isn't corrupted, does it need anything from the sysadmin to get back to work ?

  • If the node crashes and is dead : let's imagine a network failure of the NAS which corrupted the filesystem, an explosion of the datacenter where this node is, or anything that would make the node DEAD and unable to boot... Can you just setup a new node, add it to the your "cluster" and it will replicates everything on its own ?

What does the sysadmin need to do to make a node back to work (time, complexity etc..) ?

Is there any docs which covers these cases of use ?

user9517
  • 115,471
  • 20
  • 215
  • 297
will
  • 13
  • 1

1 Answers1

2

To the first point, all the datastores support replication (http://appscale.cs.ucsb.edu/datastores.html). So long as all the replicas have not gone down, you should be fine. For HBase and Hypertable if you're head node goes down the data will still be accessible for clients which have cached the location of the tablet servers.

Adding nodes is there to a certain extent with the use of Neptune (http://www.neptune-lang.org/), but adding this capacity to a far more automated extent is still needed and is on our roadmap.

I would recommend emailing the mailing list for any additional AppScale questions where you'll get a faster response: http://groups.google.com/group/appscale_community

user9517
  • 115,471
  • 20
  • 215
  • 297
Navraj Chohan
  • 36
  • 1