The utility for checking the capacity of Mysql auto-increment fields was deployed on all production servers (it is configured automatically with auto-discovery). It is a quite unique monitoring that allows you to discover potential problems with the database on the Alerta dashboard. It happens that developers, designing the database scheme, choose too small a field type, for example INT, instead of BIGINT, for a field with an auto-increment. At some point, the field is filled full, it is no longer possible to write new data, the site crashes. The utility allows you to find such databases/tables where filling of the auto-increment field is above the specified percentage.
Added the ability to set arbitrary descriptions of logrotate configs in the app state.
Changed the behavior of the deploy_only pillar in the app state. Earlier, during the deployment, a list of names was sent to this pillar, now only one name is searched for by an exact match. The previous behavior sometimes led to ambiguity, it was possible to deploy the apps we don’t want to.
In the salt-ssh-deploy repository, a script for deploying app.mini was added, and the corresponding state was added to the formula. app.mini is more convenient to use in application deployment pipelines without calling a full long deployment, it calls only php/python/ruby/static deployment states.
Added monitoring of NVME, CPU overheating and general hardware errors to our ecosystem: they are searched in dmesg. This monitoring immediately made it possible to identify several problematic servers. For example, in one of the PXC clusters there were complaints about a sharp increase in timeouts when working with the base under high loads. As a result, it turned out that some of the newly installed NVME drives had a very large spread of latency that showed itself under load. Problem analysis was done using PMM – thanks Percona for this great tool. Replacing disks and cables in the servers solved these problems and the cluster started working without delays. The new monitoring will allow such problems to be detected more quickly.
Added option to install helm to kubernetes.tools state.
Fixed a bug in date formats in alerta installation recommended pillars. The default sorting has also been changed: now the alerts are more conveniently located on the dashboard and do not constantly change their position. The color scheme has also been changed: now paler colors are used, which are less irritating to the eye.
Started to implement helm-assistant in client-side pipelines for the deployments in Kubernetes. This tool allows you to display messages that were previously visible only in the Kubernetes logs when deploying helm charts, directly in the gitlab job logs. Which makes the deployment process more transparent for both devops engineers and developers.
Added the docker-ce.compose state, which can install the latest version, instead of using the pkg state.
In the gitlab, pmm states, the installation of nginx reload crons was added, which was missing there for the correct use of renewed https certificates.
A regular accounting job has been launched, which updates the client’s project Wiki with the information about assets etc.
In accounting, now it is possible to disable the monthly license fee for the selected asset. This made it possible to get rid of many unnecessary duplicative tariffs and simplify accounting.
On the contact page of our website, links to our official contact methods have been added:
A new Microdevops architecture diagram has been published. Once you’ve tried the excellent diagramming tool Excalidraw, which is also open source, you’ll never want to use anything else. The diagrams are very vivid and clear. Many thanks to the authors of Excalidraw for this tool.
We have finally published a new main site page with rates for our main asset types, plan comparisons, user reviews and sliders, including a 🐈cat🐈 slider. As you can see, cats are not the least important characters in our ecosystem 😼.
And finally, for the first time in our history, we are starting to actively look for new clients. It may seem strange, but before that, for many years we did not look for clients, they found us themselves thanks to buzz marketing. But, since our solutions are becoming more mature and convenient, they are worth to be spread on the planet more actively 😻.
