StarCluster - Mailing List Archive

StarCluster 0.92 Released!

From: Justin Riley <no email>
Date: Tue, 18 Oct 2011 13:59:03 -0400

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi All,

StarCluster 0.92 has officially been released:

http://pypi.python.org/pypi/StarCluster
http://web.mit.edu/stardev/cluster/download.html

You can find the new docs for 0.92 here:

http://web.mit.edu/stardev/cluster/docs/0.92

Please take some time to skim through the docs as new content has been
added and old content updated/reworked. Also, the "Overview" page in
the docs now has a nice summary of StarCluster's core commands and
features. I've attached a list of highlights for the new release to
the end of this email.

Please upgrade at your leisure, test the new release candidate, and
submit any bug reports preferably to StarCluster's github issue
tracker[1] or to the mailing list[2].

Enjoy!

~Justin

[1] https://github.com/jtriley/StarCluster/issues
[2] http://web.mit.edu/stardev/cluster/mailinglist.html

- -------------
Release Notes
- -------------

Below is a list of highlights from the 0.92 release:

* StarCluster now builds a crash report that you should use when
reporting bugs. When a crash occurs StarCluster will dump a full log
for the session in ~/.starcluster/crash-report-<PID>.txt where <PID>
is the process id of the starcluster session that crashed. This
contains a full debug log for the session as well as the stack-trace
for the bug. Don't forget to remove any sensitive data from the crash
report before sending.

* StarCluster now launches the master node of a spot cluster as a
flat-rate instance instead of a spot instance, by default, for
stability. If you wish to force using a spot instance for the master
(risky) use the --force-spot-master option:

$ starcluster start -b 0.50 --force-spot-master mycluster

The only exception to this rule is if the cluster size is 1. In this
case the master will be launched as a spot instance given that it's
the only instance being requested. This is mostly for convenience so
that users can launch a single spot node for creating new AMIs, cheap
testing, etc.

* You can now easily copy files to and from a StarCluster using the
new "put" and "get" commands:

$ starcluster put /path/to/file/or/dir /remote/path
$ starcluster get /path/to/remote/file/or/dir /local/path

These commands handle recursively copying directories automatically if
necessary. For now these commands are experimental and require
enable_experimental=True to be set in the config

* The plot option (-p) to StarCluster's "spothistory" command no
longer requires matplotlib and instead renders the plot using an
embedded web server and the interactive javascript plotting library flot:

$ starcluster spothistory cg1.4xlarge -p

You can see an example plot in the new "Overview" page in the docs:

http://web.mit.edu/stardev/cluster/docs/latest/overview.html#reduce-costs-using-spot-instances

You can change the web browser launched by StarCluster to view the
spot history plot using the WEB_BROWSER setting in the [global]
section of the config.

* The "addnode" command now supports adding more than one node:

$ starcluster addnode -n 3 mycluster

You can also provide your own custom aliases for multiple nodes:

$ starcluster addnode -n 3 -a webnode,datanode,sqlnode mycluster

* The "removenode" command has a new "-k" option that allows you to
remove the node from the cluster but not terminate it:

$ starcluster removenode mycluster -k node001

You can then use the new "-x" option to the addnode command to re-add
the existing node:

$ starcluster addnode mycluster -x -a node001

This is useful, for example, when testing a plugin's on_add_node or
on_remove_node code without having to waste instance hours terminating
and creating new nodes repeatedly.

* This was mentioned in the last release candidate but there is now
"createkey", "removekey", and "listkeypairs" commands for managing
your EC2 keypairs.

* The node.ssh.execute_async() method has been updated to use "nohup"
and will now successfully leave a long running command running on the
cluster after StarCluster exits. This means you no longer need to use
'GNU screen' when trying to put a long running command in the
background in a StarCluster plugin; just use node.ssh.execute_async
instead.

** Fixed the "InvalidGroup.InUse" bug when terminating a cluster

** Fixed the "eventual consistency" bug that sometimes occurred when
StarCluster attempted to access AWS meta-data on a node too quickly
after creating it. This manifested itself as an
"InvalidInstanceID.NotFound" bug when creating a new cluster.
StarCluster now retries the failing operations up to 5 times before
throwing an error.

** Fixed a bug with when attempting to restrict SSH (port 22) access
to the cluster via a [permission] section in the config.

** Fixed a bug when defining two [volume] sections in the config that
use different partitions on the same volume.

** Fixed support for kwargs in a plugin's __init__

** Added support for IPython 0.11 in the development shell ($
starcluster shell)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk6dvmcACgkQ4llAkMfDcrn2qQCfXCavHE+Iru0lC3E12jAQl+yQ
1fIAoIvyxkKQHynJpdXm5dDrTR8sq04+
=29F7
-----END PGP SIGNATURE-----
Received on Tue Oct 18 2011 - 13:59:06 EDT
This archive was generated by hypermail 2.3.0.

Search:

Sort all by:

Date

Month

Thread

Author

Subject