Phillip B Oldham's questions

I'm having a hard time configuring Jenkins, running as a Windows Service, to connect to a git server (gitosis) and clone a repository. Jenkins is set to use the standard Git plugin.

Everything is hosted by us: the Jenkins server is on win.foo.com and gitosis is running on a linux server (git.foo.com).

If I RDP into win.foo.com I can clone (git.exe clone --progress -o origin [email protected]:myproject.git myproject) from the command line without issue.

Jenkins is running as a service and because it's not running under the same user account I've copied the .ssh folder to %SystemRoot%\System32\config\systemprofile (as defined in the %USERPROFILE% environment variable) and C:\Program Files (x86)\Git\, but neither of these seem to be picked up by Jenkins, even after a full restart.

I've tried using both git.exe and git.cmd; no joy in either case.

Jenkins is failing with the following error message:

ERROR: Error cloning remote repo 'origin' : Could not clone [email protected]:myproject.git
ERROR: Cause: Error performing command: C:\Program Files (x86)\Git\cmd\git.cmd clone --progress -o origin [email protected]:myproject.git C:\Program Files (x86)\Jenkins\jobs\myproject\workspace
Command "C:\Program Files (x86)\Git\cmd\git.cmd clone --progress -o origin [email protected]:myproject.git C:\Program Files (x86)\Jenkins\jobs\myproject\workspace" returned status code 128: Cloning into C:\Program Files (x86)\Jenkins\jobs\myproject\workspace...
Access denied
fatal: The remote end hung up unexpectedly

I'm sure I've missed something, but I'm not sure what.

Advice?

We have a linux server set up with a number of samba shares on our mixed windows/mac/linux network. The shares are accessible if we go to \\192.168.x.x\share, however we can't access the server by it's netbios name (as set-up in the smb.conf file). It's also not listed on the network, whereas all the other machines are -- including the other linux machines.

We're unable to see any difference in the config of the other machines compared to this particular one, other than that they are desktop Ubuntu machines and this is a "headless" CentOS server.

Here's the config file:

[global]
  workgroup = WORKGROUP # correct workgroup name for the network
  netbios name = foobar # same as the hostname of the box
  security = share
  guest account = nobody
  local master = no # tried yes, no difference
  preferred master = no # tried yes, no difference
  ; wins support = yes
  ; dns proxy = yes

# shares are assigned here...

The netbios name is set to the same value as the linux hostname for the box, but it's still not showing on the network whether we set the netbios name in samba or not.

What else do I need to set so that the machine shows up on the network?

How can I re-map one hostname to another, similar to how one would re-map a hostname to an IP in /etc/hosts?

For instance, how could I map oneheckofalongdomainname.com to a.com on my local machine?

Adding an entry into /etc/hosts won't work as it maps hostnames to IP addresses, whereas I need to map one hostname to another. I'm also unable to use the target domain's IP because it's subject to change.

We are expanding our network of webservers on EC2 to a number of different regions and currently use master/slave replication. We've found that over the past couple of months our slave has stopped replicating a number of times which required us to clear the db and initialise the replication again.

As we're now looking to have servers in 3 different regions we're a little concerned about these MySQL replication errors. We believe they're due to auto_increment values, so we're considering a number of approaches to quell these errors and stabilise replication:

1. Multi-Master replication; 3 masters (one in each region), with the relevant auto_increment offsets, regularly backing up to S3. Or,
2. MySQL-Cluster; 3 nodes (one in each region) with a separate management node which will also aggregate logs and statistics.

After investigating it seems they both have down-sides (replication errors for the former, performance issues for the latter).

We believe the cluster approach would allow us to manage and add new nodes more easily than the Multi-Master route, and would reduce/eliminate the replication issues we're currently seeing. But performance is a priority.

Are the performance issues of MySQL-Cluster as bad as people say?

By default when booting an m1.xlarge EC2 instance of an OpenSolaris image you're given 1.6T of drive space across 4 ephemeral devices. This is auto-configured as follows:

~# zpool status

    NAME        STATE     READ WRITE CKSUM
    mnt         ONLINE       0     0     0
      c7d1p0    ONLINE       0     0     0
      c7d2p0    ONLINE       0     0     0
      c7d3p0    ONLINE       0     0     0
      c7d4p0    ONLINE       0     0     0

What I'd like to do is to change this so that on boot the disk structure is something like the following:

    NAME        STATE     READ WRITE CKSUM
    mnt         ONLINE       0     0     0
      raidz1    ONLINE       0     0     0
        c7d1p0  ONLINE       0     0     0
        c7d2p0  ONLINE       0     0     0
    logs        ONLINE       0     0     0
      c7d3p0    ONLINE       0     0     0
    cache
      c7d4p0    ONLINE       0     0     0

... onto which I will, on boot, load data from an S3 store.

If I create the above structure and then re-image the machine, subsequent boots from this new AMI fail, either silently (as in, terminating before booting successfully) or fail to respond (once booted they I can't access via SSH or by any other means). Console output is empty in both these instances, except occasionally when there's a complaint about the devices.

Is what I'm trying to achieve possible? I'm assuming I'm simply missing the correct --block-device-mapping argument when using ec2-bundle-image but there's very little information on Google relating to this subject.

Are there any AWS or third-party services that can monitor a number of EC2 instances and restart them should one terminate unexpectedly?

We have a project which needs to take advantage of some ZFS features (snapshots, streaming, etc) but we're a little concerned with the recent events with Oracle and OpenSolaris. Is it "safe" to use the current OpenSolaris images on EC2?

We're considering whether it would be "safe" to use OpenSolaris in production on EC2 until an alternative BSD or Linux distro with native ZFS support becomes available on EC2. I understand that there are a number of OSol clones in the works, and that FreeBSD may become available soon on EC2.

By "safe" I mean that the current OpenSolaris AMIs on Amazon are stable enough for production use and that there's little/no chance Amazon will pull those AMIs in the near future.

We would like to restrict access to users within a group (Staff) so that they can only see/chat with users from another group (Support) and not each other. We would also like to ensure that users from the Support group can see all those in the Staff group.

Can this be achieved and if so, how?

We have an nginx server which proxies to a standard set-up of Apache+SVN. The nginx set-up is a very simple proxy:

server {
    server_name svn.ourdomain.tld;
    location / {
        proxy_pass http://localhost:8080;
    }
}

Apache is set-up as follows:

<Location />
    DAV svn
    SVNParentPath /var/svn
    AuthType Basic
    AuthName "Authentication Required"
    AuthUserFile /var/svn/.auth
    Require valid-user
</Location>

...which allows us to access repositories using something like http://svn.ourdomain.tld/repo. We've been running this set-up now for about 2 years without issue.

Recently we've found that we need to check out one of the repositories onto the server itself, however whenever we do so it seems to break the repo. From that point on, it will only respond with a 301 Moved Permanently error.

We've tried:

• svn co file:///path/to/repo
• svn co svn://localhost/repo
• svn co svn://svn.ourdomain.tld/repo
• svn co svn+ssh://localhost/repo
• svn co svn+ssh://svn.ourdomain.tld/repo
• svn co http://localhost/repo
• svn co http://svn.ourdomain.tld/repo

Also tried bypassing nginx, and get the same error:

• svn co http://localhost:8080/repo
• svn co http://svn.ourdomain.tld:8080/repo

Checking out from a different machine works as expected until we attempt to check out on the server, after that it refuses with the same 301 error.

What is more confusing is that this repository server also hosts our HudsonCI server, which can pulls and builds our projects hourly. This leads us to suspect that it's the svn client which is causing an error in communication.

Its also very confusing that removing then re-creating the repo using svnadmin doesn't reset the error - the repo is still unavailable even though it's "new"! Restarting apache and subversion (svnserve) has no effect on this, or the original error.

Version information:

• OS: 64-bit CentOS 4.2, 2.6.27 kernel
• svn client: 1.4.2 (same for both server and remote clients)
• svn server: 1.4.2
• httpd: 2.2.3

UPDATE:

This also happens with svn export when run on the repo server. Ran from any other box/client, there isn't a problem. Here's the workflow, to help clarify the error:

1. [~repo-server~]# svnadmin create {repo}; chown -Rf www:www {repo}
2. [remote-client]# svn checkout http://svn.ourdomain.tld/repo
3. [remote-client]# svn add file; svn ci -m ''
4. [~repo-server~]# cd /var/www; svn export file:///path/to/repo/trunk ourproject
5. [remote-client]# svn update fails with 301 error

I can also confirm that the hostname of the box doesn't have an effect here, which is very odd: whether or not svn.ourdomain.tld is added to /etc/hosts it still breaks - we thought it could be an issue with localhost routing, but that doesn't seem to be the case.

Are we missing something in the documentation which states you can't checkout a repo when the server is on the same box? How can we stop the repos becoming corrupt when we checkout locally?

Our email host only allows a small number of connections per IMAP mailbox, however we need 2x the limit number of users to be able to access the account to retrieve mails. Sending isn't required, it's simply reviewing the received mail and deleting/moving to a sub-folder.

Are there any tools available which can sit on a linux server or a Mac desktop and act as a gateway/proxy, so that our host only sees a single connection?

I'm trying to compile Python (2.6.5) on a 64bit system, using the following options:

./configure --prefix=/usr --libdir=/usr/lib64

however, libpython isn't being installed in /usr/lib64, but rather /usr/lib. This means when I invoke python from the commandline an error is raised.

I know I can hack round the problem by adding a symlink to the .so file in /usr/lib but that's not really fixing, nor understanding, the problem.

What do I need to do to get Python to install correctly?

We're in the process of setting up a continuous-integration server (Hudson running on CentOS5), and some of our unit tests will cause email to be sent. We'd like to configure the server so that all email sent via the local SMTP server is routed to a single (development) email address (rather than a user on the box) so that we can ensure the email structure/content is correct.

Can you suggest a simple SMTP server app we can install and provide a sample config to achieve this?

Are there any major security implications to bundling a private AMI which contains the private key/certificate & environment variables?

For resiliency I'm creating an EC2 image which should be able to boot and configure itself without any intervention. After boot it will attempt to:

• Attach & mount specific EBS volume(s)
• Associate a specific Elastic IP
• Start issuing backups of the EBS volume(s) to S3

However, to do this it will need the private key/pem files and will need certain environment variables to be available on start-up. Since this is a private AMI I'm wondering if it will be "safe" to store these variables/files directly in the image so that I don't need to specify any user-data information and can therefore start a new instance remotely (from my iPhone, if needed) should the instance be terminated for any reason.

I'd like to use an OpenSolaris EC2 instance as a DB server, using an external EBS volume connected to a ZFS pool.

If I set-up the ZFS pools then bundle-up the EC2 image, could I terminate the instance then safely re-attach the existing EBS volume (with data) to a new EC2 instance after it boots?

My main concern is if the instance terminates unexpectedly. I would ideally like to bring up a new instance and automate the re-attachment of the EBS volume but I'm not sure whether I could store the AMI with the pool structure already set, or whether I'd have to issue some commands at start-up to build the pool structure after attaching the volume.

I'm setting up a database server on EC2, and I need to ensure that an EBS volume is automatically attached and is available before the database service starts up.

I'm using SMF so I can test whether a particular filesystem is available before starting the db service, so there's no problem from that perspective, however I'm not quite sure how to tell the server to auto-attach the EBS volume during/after boot.

What would be the best strategy for this?

We have a small office with a large number of wireless devices (a mixture of desktop machines, laptops, and wifi-enabled phones) all working from a single Apple Airport Extreme which extends our wired network. I've added another Airport Extreme for resiliency, since we've been seeing a decrease in performance and (as far as I understand) access points can only handle a small number of clients.

I set the new AP to extend the current network so that the clients weren't constantly switching between different wireless networks, however as soon as this AP was configured all the wireless devices started seeing network trouble, flicking on and off.

I'm assuming that this is because both APs are reasonably strong, and the client can't decide which to use.

What is the best route to follow to resolve this? What I'm aiming for is wireless resiliency; preferably having two APs share the network load, or if this isn't an option then having a primary AP with a "fail-over", should the primary go down for any reason.

I'm trying to add a custom service to SMF's configuration, which seems successful in that the service starts and there is a log file, but therein lies the problem; the service, on start-up, prints some logging messages to the stderr. It seems that SMF is seeing those messages and, believing them to be errors, restarts the service, giving up after a number of tries and leaving the service off.

Here's part of the log output:

[ Mar 30 14:59:54 Enabled. ]
[ Mar 30 14:59:54 Executing start method ("java server.CustomServer"). ]
Starting server...
[ Mar 30 15:00:04 Method or service exit timed out.  Killing contract 107. ]

Running the server directly on the commandline is fine, and AFACS there are no errors being encountered during startup, other than the output.

What would be the best way to manage this service with SMF? The logging is needed for diagnosing problems, and would be problematic to disable. Is it possible to configure this service to only restart if the service exists?

Which is faster for read access on an EC2 instance; the "local" drive or an attached EBS volume?

I have some data that needs to be persisted so have placed this on an EBS volume. I'm using OpenSolaris, so this volume has been attached as a ZFS pool. However, I have a large chunk of EC2 disk space that's going to go unused, so I'm considering re-purposing this as a ZFS cache volume but I don't want to do this if the disk access is going to be slower than that of the EBS volume as it would potentially have a detrimental effect.

I have 3 identical Dell Vostro 200 slim desktop machines running Vista on the same wifi network. One of the machines has developed an issue where its wifi connection is very slow; webpage loading times are easily a factor of 10 slower than a few weeks ago.

For example: loading thesimpsons.com takes <20s on machines A and B, but machine C still hasn't completed the loading after 2mins.

I've checked the settings for the built-in wifi card and they're again identical.

Viewing the network usage while a browser open shows the download happening, and closing the browser immediately stops any download traffic - so there's nothing on the machine leeching bandwidth.

None of the machines have QoS enabled or even installed.

Vista notes the card as a Broadcom 802.11g Network Adapter "BCM43XG_NT60".

There's no previous driver to roll back, and there are no new drivers from windows update.

What would be the next step to debug this issue?

I have an hourly cron script which take some output (a mysql dump), pipes it through gzip, and aims to overwrite a file of the same name. When I run it manually as root the file is overwritten. When it is ran by the cron daemon the filename has ".1" appended to it. This keeps happening so that after a while I have lots of files like so:

myfile.gz
myfile.gz.1
myfile.gz.1.1
myfile.gz.1.1.1
myfile.gz.1.1.1.1

and so on.

ps aux|grep crond shows that the daemon is being run as root.

I've tried:

• renaming the original file, pushing the output, then removing the old file on completion, and
• deleting the original file before piping the output

but neither works as expected and I just get .1.1.1.1 files.

Script looks like this (nothing special) and is located on a CentOS box in /etc/cron.hourly:

#!/bin/bash
DATE=`date +%H`
DIR="/abs/path/to/dir"
FILE="hourly-${DATE}.gz"
OPTS="..."
mysqldump $OPTS | gzip -9 > $DIR/$FILE

Can anyone advise as to why this simple operation isn't running as expected?