Monday 28 September 2015

Tricky: docker, oraclelinux, ubuntu and ulimit -Hn

Virtualization is said to bee good for training and development, but  it should be kept out from productive environments. Some people, to whom I belong, meanwhile changed their minds and are convinced, that e.g. docker containers running ubuntu or oraclelinux, breed85/oracle12c or alike will play an increasing role in data virtualization. See also oaktable-world-agenda-2015/ To get experience and clarify certain questions we have created a Docker Project on github. This story here came up with the number of open files in a system (ulimit -Hn), crucial for productive systems. The issue discussion serves best to  understand the impact and the trickyness of the question. For whom it may concern, it is published here.

Wednesday 23 September 2015

Docker & oracle-rdbms-server-12cR1-preinstall

There is an extremely annoying  behavior of oracle-rdbms-server-12cR1-preinstall, when you want to use it in conjunction with docker images like breed85 or oraclelinux: this official package of Oracle's blocks every attempt to switch to the user oracle (su - oracle), which one is forced to do, when he wants to install Oracle SW and create Oracle databases in a Docker container. I have written 2 posts in July about that and none of them was correct. After 2 months I was forced to come back in order to accomplish our Dockerproject . Here is, what I have found out.

Saturday 19 September 2015

How to use AWR and ADDM - very short explanation

In the last post about setting up stress tests on Oracle databases I recommended AWR and ADDM to see the results of your tests. These tools are unique gems of programming. For those who are not familiar with them, here is a very short, but hopefully efficient instruction, how to handle it manually.

Friday 18 September 2015

How to setup a stress test on Oracle or equivalent databases

In a recent workshop on Oracle12c administration, I noticed that the participants had no idea how to let Oracle create archived redo logfles, i.e. they did not know how to create a relevant workload. In the exercises of the workshops you can find one, which performs some workload on CPU and I/O, but nothing about writing so much to the online redo logs, that the log writer process LGWR needs to copy the contents to the archived redo log files.
In summary, the students - all computer scientists working several years in IT companies - had never seen a stress test before. To solve the issue, I passed an old script of mine to them. To whom it may concern, I will publish it here, because it is really simple.