Operations and Administration - for SaaS Mode¶
The last thursday of each month, the latest release of the product is deployed.
This release includes all the new features and every improvements developed during the past month.
To be be deployed and qualified as stable release, it also needs to pass all our quality control processes (such as our extended user acceptance tests or the container analysis).
Please also note, in case of security issue we would deploy a patched release as soon as we have fixed it.
Details about our releases content are published on the dedicated page.
All deployments are automated thanks to our Ansible playbooks.
All our playbooks are versioned, maintained and reviewed by the Toucan Toco tech team.
As a best practice, we never directly connect to a Toucan Toco node in order to run manual commands. We reduce the risk of human error and deployment is able to auto-scale.
All logs generated by Toucan Toco’s applications are centralized in an Elastic stack. Toucan Toco’s tech team can follow activity on these apps as well as any warnings and errors thanks to Kibana dashboards.
We also have centralized system logs like syslog, auth, nginx (access and error) and fail2ban. We are able to detect brute force attacks, spam and malicious behavior on our dashboards. For each detected pattern, we receive automated alerts thanks to Elastalert.
Our log rentention policy is about 8 weeks long in our Elastic stack but we keep - by default - 14 weeks of web access/error logs and 52 weeks of app logs in our servers.
As a best practice, we never need to directly connect to a Toucan Toco node to follow activity and logs. This is at the core of our ability to scale our monitoring.
We regularly and automatically scan our servers in search for:
- open ports
- lack of security updates
We ensure having an up-to-date environment (system, security, patches…).
Our monitoring services alert us when:
- a server becomes unresponsive
- a server shows unusual CPU, memory or disk activity
- a server is getting closer to its hardware limits
- an application status page shows that it’s not OK
- one of the following ports is not listening: 443/80/22
Furthermore, we use OSSEC to alert us to possible intrusions.
This monitoring runs 24/7 and every alert is checked to ensure a fast reaction from the Toucan Toco tech team.
Every week these services send us detailed performance and uptime reports.
These regular reports help us to identify pontential regressions or bottlenecks that can then be fixed.
Watch and patch management¶
To discover new vulnerabilities and patch against them as quickly as possible, we follow:
|Database||MongoDB||Mongo CVE DAdministration, exploitation & internal security details|
|Database||MongoDB||Mongo Security Checklist|
|Application||Python||Python CvAdministration, exploitation & internal security Database|
|Container||Docker||Docker Dev Mailing list|
|Container||Docker||Docker User Mailing list|
|System||Ubuntu||Ubuntu LTS packages|
|System||Ubuntu||Ubuntu Security List|
|System||Debian||Debian Security List|
And Github’s issues/announces of main projects we use.
As soon a security patch is available, we automatically applied it to our whole infrastructure by using our Ansible scripts.
Otherwise, our infrastructure is fully updated every 2 months with our Ansible playbooks. But before applying updates everywhere, we use a staging node to be sure there will be no regression.
This update processes can very occasionally lead to a short downtime that we do out of office hours.
If the infrastructure or the applications are impacted by a known vulnerability, we always send a mail report to the client to warn and explain how we recover it.
We run a daily backup process for each instance/project.
The backup is a complete snapshot which is encrypted by a GPG key
(dedicated to the instance/project) and exported over
our exclusive backup nodes.
GPG keys are only available to the Toucan Toco’s admins and stored in our passwords manager system.
All the backups are exported to our dedicated storage servers hosted by OVH in another region.
By default we keep a retention of 20 daily backups for each instance/project.
We also regularly test and challenge our backup and restoration scripts.
Restoring an instance or a project is a fully automated and fast process.
By culture, we keep a logbook of every issues on the infrastructure.
Each logbook entry describes:
- what’s going on
- how did we understand the issue
- what did we do to solve the problem
- what are the impacts
- what do we need to do to avoid it next times
The logbook is open to every Toucan Toco employees. The knowledge, about the life and the issues on the infrastructure, is shared and maintained by every one.
Communication during issues¶
As soon as we detect an issue, your dedicated account manager and/or client success manager will contact you to explain the issue, the potential impacts and give you an estimated resolution time.
When the issue is closed, you can expect a post-mortem report, mainly extracted from our logbook (cf previous paragraph), with details about the investigation and the resolution process.
This emergency communication is available 24x7.
Our main support channels are emails via firstname.lastname@example.org.
This support is open between 9:00 and 18:00 (Paris hour) during the working days.
On-Call duty team¶
Project instance and server decomission¶
Each time we need to decomission a project instance:
- the dedicated stack is shutdown (virtual hosts, API process, workers, queue server, database)
- all data, logs and associated configuration are erased
Each time we need to decomission a server:
- data and home partition are fully formatted
- we force a basic rewrite of the partition (with a basic
ddcommand), thus no block could be restored from their previous state
- then we release the server to Scaleway.
A decomissioned server is always left without any data.
We have exactly the same approach for any SAN or storage volumes.
Container runtime security monitoring¶
To ensure that our customer’s instances are not compromised while running and detect any suspicious behavior which can lead to security issue, Toucan uses Falco on its infrastructure to monitor running containers and hosts in real-time.
Whenever Falco detects a scenario that is not on the Toucan team’s whitelist, it sends an alert to the team.
Here are some examples of suspicious activities:
- RCE (Remote Code Execution) inside a container.
- Package installation during the runtime of a container.
- Shell binding to a suspicious file descriptor.
- Netcat Program runs inside a container that allows remote code execution.