jl6's questions

I have a pair of 3TB drives in a btrfs filesystem, and I wish to expand this filesystem by adding two more 2TB drives, using the btrfs device add command. Data in the current filesystem is mirrored in raid1 mode.

I am hoping the resulting filesystem will have 5TB usable capacity with all data in raid1 mode.

But, the official wiki states:

When you have drives with differing sizes and want to use the full capacity of each drive, you have to use the single profile for the data blocks, rather than raid0.

But, this answer suggests that there may have been a patch to use size-mismatched devices to full capacity with full data mirroring:

There is a quasi-round-robin patch in the pipeline to improve on this. Of course, it is still impossible to pair all chunks on different devices if you have a 500GB and a 2TB disk. The patch is intended more for situations like 1 x 1TB + 2 x 500GB where each small disk should prefer to mirror/stripe with the big disk instead of the other small disk.

Which is correct?

I have multiple domains (example1.com, example2.com, ...) hosted on the same IP. One of these domains (example3.com) has an SSL certificate and I want to serve it over HTTPS, while keeping all the other sites on HTTP. I have this set up and working correctly, so far.

My problem is that requests to https://www.example1.com (note the s) are getting handled by the apache configuration section for example3.com (which begins <VirtualHost *:443>), which causes problems as this is a Django site, and amongst other things it generates a 400 Bad Request due to example1.com not being in Django's ALLOWED_HOSTS setting.

I understand that the nature of SSL means that the Host: header is not known until after the secure connection is established. But is there a way of getting apache to reject any requests over HTTPS that are not for example3.com?

I had expected use of the ServerName directive within the SSL Virtualhost to restrict that section to just the named host, but upon closer inspection of the docs it seems that is only the case for name-based virtual hosts.

Edit: I have tried adding a catch-all default as the first section, like the following:

<VirtualHost *:443>
  ServerName default.only
  <Location />
    Require all denied
  </Location>
</VirtualHost>

This causes the following error:

[Thu Dec 04 10:31:27.922801 2014] [mpm_event:notice] [pid 10498:tid 3074255488] AH00491: caught SIGTERM, shutting down
[Thu Dec 04 10:31:29.928483 2014] [ssl:emerg] [pid 30518:tid 3074300544] AH02240: Server should be SSL-aware but has no certificate configured [Hint: SSLCertificateFile] ((null):0)
[Thu Dec 04 10:31:29.928551 2014] [ssl:emerg] [pid 30518:tid 3074300544] AH02312: Fatal error initialising mod_ssl, exiting.

In a file in /etc/rsyslog.d/ I have the following:

local3.*        /var/log/mylog.log

When I log to this facility, records have timestamps which look like this:

Apr 27 21:12:20 hostname msg

How do I set things up so that they look like this instead?

2014-04-27T21:12:20 hostname msg

Amazon, Facebook and Google must take in TBs if not PBs per day, so does that mean they have people continuously installing new hard disks and cables in new racks in the same way you might build a regular server, or is there a different technology used for connecting drives en masse?

I'm expecting the answer to be no but I can't quite reason out why not.

I have an Ubuntu server and I am interested in measuring how much data is being transferred (both in and out) across a network interface, by a specific user. How do I do it?

N.B. I know how to measure the total data transfer of the whole machine. I want to restrict my monitoring to a specific user.

I have a series of cron jobs and I would like to test them in a more timely and automated way than the following:

1. Test individual jobs in isolation by running them manually.
2. Install job and update crontab.
3. Wait for cron to execute jobs.
4. Observe results.

I'm thinking that running a test script as root (in a development environment, of course) and fiddling with the system clock might be beneficial. What is best practice?

I would like to have cron output sent to an external email address, via a minimal SMTP client that sends via Gmail. I think I need this client to support TLS and /etc/aliases, and I would like configuration to be as simple as possible.

acct logs execution of all processes, and the users that execute them, recording statistics such as real time elapsed, and CPU time elapsed. I have seen it suggested that this might be useful in a forensic context, i.e. working out who executed what, and when. But as it just records the name of the process, I'm wondering what value it really adds.

If I execute

$ cp /usr/bin/cc vi
$ ./vi malware.c -o ls
$ ./ls

then the process accounting log will only contain entries named "cp", "vi", and "ls" - all innocuous.

Therefore process accounting appearas to offer limited security benefits. Any contrary opinions?

I have a server build script which uses apt-get to install packages. It then puts pre-written configuration files directly in place, so the interactive post-install configuration dialog in packages such as postfix is not needed. How do I skip this stage of the installation? It creates a piece of manual intervention that I would rather avoid.

I am aware of the -qq option, but the manpage warns against using it without specifying a no-action modifier. I do want to perform an action, I just want to suppress a specific part of it.

I've got about 200GB of data that I'd like to serve on a low-volume web site. I've currently got this on shared hosting at Dreamhost, but I'd like to upgrade to something that gives me root access like Amazon EC2. Although running an Amazon EC2 instance 24/7 looks more expensive than the competition, $0.10/GB-month for EBS appears to be a very good rate for storage. Is there anybody offering comparable storage for less?

Thanks!