NETWAYS Blog

Attending conferences of a certain size always brings the pain of deciding what talks to listen to. I really wished I could listen to all of them live but at least there are the OSDC archives, so I will be able to watch what I’ve missed when the videos will be edited and uploaded. Following are my personal impressions of the talks I chose from the tracks.
Despite being at least a bit tired from attending the OSDC evening event at the Paulaner Bavarian restaurant in Berlin yesterday talks were interesting enough that from the first talk on day 2 of the Open Source Datacenter Conference on the two rooms where talks were given were full of attendees again.
As first talk of day 2 Timo Derstappen used the Pixar movie plot schema „Once upon a time…“ to give an introduction into microservices and how orchestrate them using the „ambassador pattern“. He showed with his demo that systemd and fleet on CoreOS can be a real alternative when running multiple containers depending on each other.
After the first coffee break of the day Ingo Friepoertner gave a short overview over the history of databases and why NoSQL databases will not replace relational databases but are the better alternative for many applications. Using the concept of „Polyglot Persistence“ you will want to use the right store for the right job, so split data into different databases: Financial data in an RDBMS, user sessions in a KeyValue store and recommendations into a graph database and so on. To avoid getting stuck in a swamp of many different databases from many different vendors you can use multi-model databases to consolidate several of the beforementioned technologies into one product.

Ingo Friepoertner with Polyglot Persistence & Multi-Model #nosql Databases in MOA 4 #OSDC /eh pic.twitter.com/YQaUJamRh7

— netways (@netways) April 23, 2015

In the next talk Matthias Klein showed us how to keep development close to production by creating your own package repositories and copy only selected packages into them. Then use vagrant or Xen and puppet to create test hosts that resemble the production hosts as close as possible to roll out new packages, run tests from Jenkins and, if all went well, deploy the same packages on production systems. One of the most important factors will be to convince everyone that only packages in your repositories are allowed and that every configuration change has to be implemented in puppet so it can be reviewed more easily and tested on test hosts before the same configuration (with just some parameters replaced) goes on production hosts.
After the great lunch John Spray showed us what CephFS is, how it developed lately and what the current state is. Being not so familiar with Ceph like I wished I was, I was glad hearing that CephFS, a distributed filesystem on the Ceph backend, is available now (even if it’s not supported for enterprise use) and CephFS volumes can be mounted simultanousely on different hosts just like a network share. A great „new“ feature (which is available for about a year now, which is almost no age for enterprise grade software) is the combination of Cache Tiering and Erasure Coding which basically means having a fast but small pool of disks (preferably SSDs) that handles all of your I/O and a second pool of big, slow disks where data is not replicated like in standard Ceph pools but split like on RAID 5 / 6 / Z systems where automatically „cold“ data, which is rarely used, is moved to.

A really cool talk about @Ceph by @jcspray. Thanks for coming over! #osdc /be pic.twitter.com/4QV7wYkxgc

— netways (@netways) April 23, 2015

Next was a talk I anticipated a lot, personally. Pere Urbon talked about one of my most favourite toolsets, the ELK stack. He gave a quick explanation about what Elasticsearch, Logstash and Kibana are and what they are used for. Then he dug a bit into some real life examples about scaling the stack, especially by using a broker like Redis or RabbitMQ. In fact, RabbitMQ might be a real alternative to Redis and I will definitely see into the different possibilities, the various brokers have to offer. Not only Redis and RabbitMQ but Apache Kafka and some others. During the talk we learned a bit about the new features of the to-be-released-today Logstash 1.5 and the coming-someday Logstash 2.0. After the talk there was only one question unanswered: How is the Green Lantern Corps logo out of the slides related to the Marvel comic multiverse, Elastic normally gets most of its names out? 😉

Pere Urbon talks about Scaling #Logstash:A Collection oft War Stories in MOA 5 #OSDC /eh pic.twitter.com/4fqeiDUZoO

— netways (@netways) April 23, 2015

The last talk of the day I attended was Bernds talk about why to use Icinga instead of Nagios. Whoever reads this blog should already know that Icinga (2) is way cooler than Nagios so I will not go into detail about his talk. But you should definitely look at the video in the archive when it’s online even if you’re already convinced because Bernd’s talks are always worth the while.

OH: "I will focus on #icinga2 cause that's our new hot shit" – @gethash at #osdc 😀

— Michael Friedrich (@dnsmichi) April 23, 2015

Thomas Widhalm

Manager Operations

---

Pronomina: er/ihm. Anrede: "Hey, Du" oder wenn's ganz förmlich sein muss "Herr". Thomas war Systemadministrator an einer österreichischen Universität und da besonders für Linux und Unix zuständig. Seit 2013 ist er bei der NETWAYS. Zuerst als Consultant, jetzt als Leiter vom Operations Team der NETWAYS Professional Services, das unter anderem zuständig ist für Support und Betriebsunterstützung. Nebenbei hat er sich noch auf alles mögliche rund um den Elastic Stack spezialisiert, schreibt und hält Schulungen und macht auch noch das eine oder andere Consulting zum Thema. Privat begeistert er sich für Outdoorausrüstung und Tarnmuster, was ihm schon mal schiefe Blicke einbringt...

Lies mehr von Thomas und triff unser Team