<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Alex Howells - Medium]]></title>
        <description><![CDATA[Random words about technology, life, travel, spaniels. Entropy as a service. - Medium]]></description>
        <link>https://medium.com/alexhowells?source=rss----e2033b405558---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Alex Howells - Medium</title>
            <link>https://medium.com/alexhowells?source=rss----e2033b405558---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:11:09 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/alexhowells" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Sales in 2016, and ‘The Slack Effect’]]></title>
            <link>https://medium.com/alexhowells/sales-in-2016-and-the-slack-effect-d9d3bc70a5f3?source=rss----e2033b405558---4</link>
            <guid isPermaLink="false">https://medium.com/p/d9d3bc70a5f3</guid>
            <category><![CDATA[slack]]></category>
            <category><![CDATA[spam]]></category>
            <category><![CDATA[sales]]></category>
            <category><![CDATA[dcos]]></category>
            <dc:creator><![CDATA[Alex Howells]]></dc:creator>
            <pubDate>Mon, 19 Sep 2016 15:56:39 GMT</pubDate>
            <atom:updated>2016-09-19T16:07:14.973Z</atom:updated>
            <cc:license>http://creativecommons.org/licenses/by/4.0/</cc:license>
            <content:encoded><![CDATA[<p>I can’t speak for anyone else, but I <strong>would</strong> be curious if this is widespread. I’ve witnessed a significant uptick in unsolicited communications (‘spam’) this year related to technology companies. Today’s example is Sysdig. They’re not the first to think of this “growth hack” (and won’t be the last).</p><p>For quite some time I’ve received <strong>many</strong> messages directed at the address which is used for LinkedIn, to the point I adapted my mail filtering, accepting messages enveloped to that address only from @linkedin.com. This is annoying but <em>almost</em> seems blessed by LinkedIn.</p><p>The most recent development, ‘The Slack Effect’, is that sales teams are harvesting addresses from Slack teams intended for community interaction. In the case of Sysdig, they harvested the <a href="http://chat.mesosphere.com/">Mesosphere DC/OS</a> participants, and the ‘lucky recipients’ started hearing from Daniel Tay on 9/15/16. Assuming you didn’t reply or opt out, another message arrived today!</p><blockquote>Hi Alex,</blockquote><blockquote>I’ve talked with a number of engineers about their plans and future strategy for containerizing their infrastructure and applications.</blockquote><blockquote>Honestly, it seems like on many levels, the industry is still figuring out these new technologies, and there are no standard best practices in place yet. Does this resonate with you? If so, could we set up a time to talk?</blockquote><blockquote>We recently released Mesos/Marathon Universe support for our open source visibility and troubleshooting tool, sysdig, and the feedback from the community has been overwhelmingly positive.</blockquote><blockquote>Do you have some time to chat? In any case, check out our public Sysdig Slack channel.</blockquote><blockquote>Cheers,</blockquote><blockquote><strong>Daniel Tay</strong></blockquote><blockquote>Product Specialist, Sysdig</blockquote><blockquote>109 Stevenson St. Suite 200 San Francisco, CA 94105</blockquote><blockquote>(415) 912–9305 | LinkedIn | Sysdig.com | Sysdig.org</blockquote><blockquote>If you’d like me to stop sending you emails, please click here</blockquote><p>It’s technically no more challenging to filter on envelope, as with LinkedIn, however with Slack the potential false positive rate is markedly higher and some legitimate communication may be lost because someone cannot ask, “What e-mail address does Alex use?” by checking a profile.</p><p>I’m not a lawyer but I’d also query whether this kind of harvesting undertaken on behalf of a U.S. company violates <a href="https://www.ftc.gov/tips-advice/business-center/guidance/can-spam-act-compliance-guide-business">CAN-SPAM</a>?</p><p><strong>Edit: </strong>To their credit, Sysdig <a href="https://twitter.com/sysdig/status/777896340520964096">responded on Twitter</a> and is investigating.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d9d3bc70a5f3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/alexhowells/sales-in-2016-and-the-slack-effect-d9d3bc70a5f3">Sales in 2016, and ‘The Slack Effect’</a> was originally published in <a href="https://medium.com/alexhowells">Alex Howells</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cost Effective Infrastructure]]></title>
            <link>https://medium.com/alexhowells/cost-effective-infrastructure-a5c010dc7bac?source=rss----e2033b405558---4</link>
            <guid isPermaLink="false">https://medium.com/p/a5c010dc7bac</guid>
            <category><![CDATA[tips-and-tricks]]></category>
            <category><![CDATA[linux]]></category>
            <category><![CDATA[oracle]]></category>
            <category><![CDATA[centos]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Alex Howells]]></dc:creator>
            <pubDate>Sat, 17 Sep 2016 15:18:27 GMT</pubDate>
            <atom:updated>2016-09-17T15:21:47.421Z</atom:updated>
            <cc:license>http://creativecommons.org/licenses/by/4.0/</cc:license>
            <content:encoded><![CDATA[<p>Alternative titles might be <em>“To OVH, or not to OVH…”</em> or <em>“Kickstart at OVH”</em>.</p><p><strong>Early Warning: </strong>I think this whole post only makes sense for <strong><em>either </em></strong>seriously cash-strapped startups trying to bootstrap, or for personal endeavours. Everyone else will read the ‘counterpoints’ and write a check to <a href="https://aws.amazon.com/">Jeff Bezos</a>. That’s okay and should not lead to <a href="https://xkcd.com/386/">this reaction</a>.</p><p>I’m an avid proponent of <a href="https://www.vultr.com/?ref=6902536">Vultr</a> for a number of reasons -</p><ul><li>Diverse set of <a href="https://www.vultr.com/locations/?ref=6902536">datacenters</a> and quick instance provisioning time. Deployment in seconds via API call, into 15+ places across the globe. Development velocity is <em>so much higher</em> due to this “simple” fact. Instances being billed by the hour keeps the prototyping cost down.</li><li>Although a bit niche, they have <a href="https://www.vultr.com/features/bgp/?ref=6902536"><em>seamless support</em></a> for BGPv4 peering, permitting me to announce IPv4 and IPv6 space and enabling interesting projects like a budget (&lt;$100/month) DNS setup using IP Anycast.</li></ul><p><strong><em>However</em></strong> with that said, there is no “one size fits all” in life or infrastructure. As the title implies, some ‘steady state’ infrastructure wants 2–4GB+ RAM (c.f. GitLab) or larger amounts of block storage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SlVEKBpA4OURBiDIP8WLAQ.png" /></figure><p>Using the ‘Storage Instances’ at Vultr is <strong><em>purportedly</em></strong> a good way to get between 125GB — 1000GB for somewhat cost effective monthly amount, unfortunately this is only available in some datacenters and my experiences were subpar — applications spent most of their time in I/O wait.</p><p>A few cheap hosting providers exist offering dedicated servers in Europe. Two of the largest are <a href="https://www.hetzner.de/us/hosting/">Hetzner</a> and <a href="https://www.ovh.com/us/">OVH</a>. In recent years they have started a <a href="https://www.ovh.com/us/news/articles/a1131.production-server-ovh-bhs">push into North America</a> to expand their addressable market. This is good.</p><p>In the case of the latter they have a number of ‘trade brands’ purporting to target specific markets from “ultra cheap” through to “small business”. Pricing with <a href="https://www.kimsufi.com/us/en/">Kimsufi</a> starts at just $6/mo for a dedicated server equipped with an Atom N2800, 2GB RAM, 500GB HDD and FastE internet connection. Alternatively with <a href="https://www.soyoustart.com/us">So You Start</a> a Xeon W3520, 16GB RAM, 2x2TB HDD would cost you $42/mo — this seems like ‘good value’.</p><p>A number of counterpoints to dedicated servers in OVH exist -</p><ul><li>Delivery of things via <em>*aaS</em> means someone else maintains them for you.</li><li>Storage durability in Amazon Web Services will likely be much higher. Big providers implement <a href="http://research.microsoft.com/en-us/um/people/chengh/papers/LRC12.pdf">Erasure Coding</a> for object storage and having data stored on many servers in this manner beats a parity-based array. The costs for ~5TB run from $35/mo (Glacier) to ~$150/mo (S3).</li></ul><p>I concluded I was happy to maintain a couple of systems. <a href="https://www.urbandictionary.com/define.php?term=YKIOK">YKIOK</a>. ;-) Dedicated systems gave me some guarantees around I/O oversubscription and the $/GB is attractive for RAM and HDD.</p><p>Approximately once a year <a href="https://spideroak.com/">Spideroak</a> runs an <a href="https://www.cnet.com/news/spideroak-offers-unlimited-cloud-storage-for-us125year/#!">unlimited promotion</a> and if you just need offsite backups, I heartily recommend their services.</p><p>I couldn’t find a one size fits all at OVH so I solved in two ways -</p><ul><li>Storage — A way to get 5TB+ usable capacity with RAID protection. <strong>Solution:</strong> <a href="https://www.ovh.com/us/dedicated-servers/details-servers-range-STORAGE-id-2016-FS-12T.xml">OVH FS-12T</a> (Xeon D-1521, 16GB RAM, 3x4TB HDD).</li><li>Hypervisor — A combination of 32GB+ RAM with solid-state storage. Would be carved up into 14 ‘slots’, each 2GB RAM and 30GB SSD. <strong>Solution:</strong> <a href="https://www.ovh.com/us/dedicated-servers/details-servers-range-HOST-id-2016-HOST-32L.xml">OVH HOST-32L</a> (Xeon D-1521, 32GB RAM, 2x480GB SSD).</li></ul><p>One upside to Vultr and Amazon Web Services is they make it <strong>very easy </strong>to bring your own operating system. In this department, OVH is <a href="https://www.ovh.co.uk/dedicated_servers/distributions/"><em>a bit lacking</em></a>. Whilst they purport to support lots, the installs are <strong>not</strong> vanilla and when configuring servers I opted to have CentOS 7.2 but there were surprises.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-WPvzakznXfUwqk-8Mm-zQ.png" /></figure><blockquote>[root@ns1234567 ~]# uname -a<br>Linux ns1234567.ip-123–123–123.eu 3.14.32-xxxx-grs-ipv6–64 #7 SMP Wed Jan 27 18:05:09 CET 2016 x86_64 x86_64 x86_64 GNU/Linux</blockquote><blockquote>[root@ns1234567 ~]# yum list installed | grep kernel<br># No output was returned to this command</blockquote><p>Immediately you notice that’s a <a href="https://grsecurity.net/">grSecurity</a> kernel and which will <strong>not</strong> be updated by <em>`yum update` </em>to receive security patches in the future. Additionally it trails the latest version (3.14.79) by a significant amount.</p><p>On top of that they remove FirewallD which is part of @core in CentOS. Systems also come with SELinux set to “disabled” (not even “permissive”).</p><blockquote>[root@ns1234567 ~]# iptables -L -v -n<br>Chain INPUT (policy ACCEPT 2677 packets, 769K bytes)<br> pkts bytes target prot opt in out source destination</blockquote><blockquote>Chain FORWARD (policy ACCEPT 0 packets, 0 bytes)<br> pkts bytes target prot opt in out source destination</blockquote><blockquote>Chain OUTPUT (policy ACCEPT 2970 packets, 448K bytes)<br> pkts bytes target prot opt in out source destination</blockquote><blockquote>[root@ns1234567 ~]# sestatus<br>SELinux status: disabled</blockquote><p>Finally there are monitoring scripts which “phone home” to OVH. Investigating /usr/local/rtm reveals these send a vast array of information back to the mothership about running processes, listening ports, et al.</p><blockquote>[root@ns1234567 ~]# grep rtm /etc/crontab<br>*/1 * * * * root /usr/local/rtm/bin/rtm 24 &gt; /dev/null 2&gt; /dev/null</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/707/1*BGYK-wATFtZuLA0SdDmnbA.png" /></figure><p>This feels suboptimal in a number of different ways but it is also difficult to argue with their pricing. So how do you fix the above?</p><p>I wanted to use an operating system which is not provided by OVH at all. Thusly the question became “Is it possible to install Oracle Linux 7.2?”. Ideally this would give me -</p><ul><li>A clean installation with all packages being ‘traceable’ to origin. Avoidance of Perl scripts in dark corners phoning home.</li><li>Usage of a modern kernel (4.0+) which will receive security fixes. Adopting Oracle Linux means you have UEK4 available!</li><li>A repeatable technique for install which works at OVH and Vultr.</li><li>Try to avoid the process being too “hands on” and time intensive. Minimise the number of human steps, to also minimise the error rate.</li><li>Systems to come up with FirewallD and SELinux out of the box. Remember: Every time you `setenforce 0`, <a href="https://medium.com/u/ea2ba511c8b">Major Hayden</a> <a href="http://stopdisablingselinux.com/">gets upset</a> and <a href="http://people.redhat.com/dwalsh/">Dan Walsh</a> weeps — please don’t make a grown man weep.</li></ul><p>I dug into how the PXE environment works at OVH and what is possible. Sadly whilst they do offer a ‘Rescue Environment’ there is no obvious way to boot from an arbitrary kernel with a set of options declared nor anything equivalent to the <a href="https://www.vultr.com/docs/ipxe-boot-feature/?ref=6902536">iPXE Boot Feature</a> at Vultr.</p><p>I resolved to use CentOS 7.2 — installed by OVH — as the “jump off” point. They’ve delivered all my systems using GRUB 2 and use its RAID support.</p><blockquote>[root@ns1234567 ~]# lsblk<br>NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT<br>sda 8:0 0 447.1G 0 disk<br>├─sda1 8:1 0 19.5G 0 part<br>│ └─md1 9:1 0 19.5G 0 raid1 /<br>├─sda2 8:2 0 427.1G 0 part<br>│ └─md2 9:2 0 427.1G 0 raid1 /home<br>└─sda3 8:3 0 511M 0 part [SWAP]<br>sdb 8:16 0 447.1G 0 disk<br>├─sdb1 8:17 0 19.5G 0 part<br>│ └─md1 9:1 0 19.5G 0 raid1 /<br>├─sdb2 8:18 0 427.1G 0 part<br>│ └─md2 9:2 0 427.1G 0 raid1 /home<br>└─sdb3 8:19 0 511M 0 part [SWAP]</blockquote><p>Installing a suitable kernel and initrd is ‘Step 1’. You can copy them over from the <a href="http://edelivery.oracle.com/linux">Oracle Linux 7.2 ISO</a> -</p><blockquote>[root@mirror pxeboot]# pwd<br>/mnt/iso/images/pxeboot<br>[root@mirror pxeboot]# ls -l<br>total 88358<br>-rw-r — r — 2 root root 39920608 Nov 26 2015 initrd.img<br>-r — r — r — 1 root root 664 Nov 26 2015 TRANS.TBL<br>-rw-r — r — 2 root root 45402044 Nov 26 2015 upgrade.img<br>-rwxr-xr-x 2 root root 5154576 Nov 20 2015 vmlinuz</blockquote><p>If you were using CentOS 7.2 then you can get the kernel here -</p><blockquote><a href="http://mirror.centos.org/centos-7/7.2.1511/os/x86_64/images/pxeboot/">http://mirror.centos.org/centos-7/7.2.1511/os/x86_64/images/pxeboot/</a></blockquote><p>After that is done ‘Step 2’ is telling GRUB 2 to boot from it -</p><pre>menuentry “Network Install” {</pre><pre>  insmod part_msdos</pre><pre>  insmod diskfilter</pre><pre>  insmod mdraid09</pre><pre>  insmod ext2</pre><pre>  linux /boot/vmlinuz net.ifnames=0 ksdevice=eth0 ip=1.1.1.1 netmask=255.255.255.0 gateway=1.1.1.254 dns=213.186.33.99 ks=http://mirror.yourdomain.tld/ovh-host32l.ks</pre><pre>  initrd /boot/initrd.img</pre><pre>}</pre><p>Things you’ll want to note -</p><ul><li>I don’t think OVH uses DHCP servers and therefore you need to ensure to configure some values (<em>ip, netmask, gateway, dns</em>) on the kernel cmdline. The right values can be found in <em>`ip addr`</em> and <em>`ip route`</em>!</li><li>When iterating on your kickstart template it can be <strong><em>helpful</em></strong> to fire up a browser and use the ‘IPMI’ tab to connect to ‘Remote KVM’ to watch. You’ll spot any issues quickly, adjust, reset and be done quickly!</li></ul><p>This all worked <strong>great</strong> and it now takes under 15 minutes from OVH delivering the ordered system to having Oracle Linux 7.2 installed.</p><p>Since this approach uses Kickstart, any EL-based operating system works. Anyone else using FS-12T or HOST-32L can consult <a href="https://gist.github.com/agh/8298baea427154fbc10fb9b7a7d5efe8">uploaded examples</a> of the kickstart configurations I use for those machine types.</p><p>In summary I’m happy with OVH now I’m not using their EL installs and will continue using Vultr for transient workloads, and IP Anycast endeavours. Hopefully this post will save someone else time if you use OVH and desperately want something “a bit different” in operating system terms.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a5c010dc7bac" width="1" height="1" alt=""><hr><p><a href="https://medium.com/alexhowells/cost-effective-infrastructure-a5c010dc7bac">Cost Effective Infrastructure</a> was originally published in <a href="https://medium.com/alexhowells">Alex Howells</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Tech Tip: Writing a Linux ISO onto USB stick using macOS Sierra]]></title>
            <link>https://medium.com/alexhowells/writing-an-iso-onto-usb-stick-using-macos-eec16d781f9?source=rss----e2033b405558---4</link>
            <guid isPermaLink="false">https://medium.com/p/eec16d781f9</guid>
            <category><![CDATA[tips-and-tricks]]></category>
            <category><![CDATA[linux]]></category>
            <category><![CDATA[mac]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Alex Howells]]></dc:creator>
            <pubDate>Sun, 28 Aug 2016 13:41:01 GMT</pubDate>
            <atom:updated>2016-09-13T09:24:12.124Z</atom:updated>
            <cc:license>http://creativecommons.org/licenses/by/4.0/</cc:license>
            <content:encoded><![CDATA[<p>For some reason I have to try hard to remember this one <strong>every time</strong> a system needs reinstalling from USB stick. Thankfully it isn’t complicated! This should work for <strong>any</strong> distribution, although the example is specific.</p><p>If you needed to get Oracle Linux 7 onto a USB stick in bootable state -</p><ul><li>Download the ISO from <a href="https://edelivery.oracle.com/linux">Oracle eDelivery</a>, or wherever else the ISOs are available if you’re doing CentOS or Ubuntu.</li><li>Perform a conversion into from ISO into UDRW format using <em>`hdiutil`</em>.</li></ul><pre>hdiutil convert -format UDRW -o oracle_linux.img ~/Downloads/oracle_linux.iso</pre><ul><li>Plug in the USB stick, and look for it in <em>`diskutil list`</em>.</li></ul><pre>scorch:~ agh$ diskutil list<br>/dev/disk0 (internal, physical):<br> #: TYPE NAME SIZE IDENTIFIER<br> 0: GUID_partition_scheme *500.3 GB disk0<br> 1: EFI EFI 209.7 MB disk0s1<br> 2: Apple_CoreStorage scorch 499.4 GB disk0s2<br> 3: Apple_Boot Recovery HD 650.0 MB disk0s3</pre><pre>/dev/disk1 (internal, virtual):<br> #: TYPE NAME SIZE IDENTIFIER<br> 0: scorch +499.1 GB disk1<br> Logical Volume on disk0s2<br> 9F7F75F3–41C4–482C-8396–9602AA2430D8<br> Unlocked Encrypted</pre><pre>/dev/disk2 (external, physical):<br> #: TYPE NAME SIZE IDENTIFIER<br> 0: FDisk_partition_scheme *32.0 GB disk2<br> 1: 0xEF 6.4 MB disk2s2</pre><ul><li>Unmount the USB stick so we can write onto it without interference.</li></ul><pre>scorch:~ agh$ diskutil unmountDisk /dev/disk2<br>Unmount of all volumes on disk2 was successful</pre><ul><li>Write the converted image to the disk using <em>`dd`</em>.</li></ul><pre>scorch:~ agh$ time sudo dd if=oracle_linux.img of=/dev/rdisk2 bs=1m<br>Password:<br>378+1 records in<br>378+1 records out<br>396974080 bytes transferred in 160.293913 secs (2476539 bytes/sec)</pre><pre>real    2m43.577s<br>user    0m0.009s<br>sys     0m3.967s</pre><p>That’s it. Finis. You probably need to eject the disk in macOS and can use it. Boot the system you need to install with the USB stick inserted and go nuts. Note: On a HP ProLiant you’d hit F11 during POST to get the Boot Menu.</p><p>This probably works on other versions of macOS but I tested on Sierra!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eec16d781f9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/alexhowells/writing-an-iso-onto-usb-stick-using-macos-eec16d781f9">Tech Tip: Writing a Linux ISO onto USB stick using macOS Sierra</a> was originally published in <a href="https://medium.com/alexhowells">Alex Howells</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>