W3C home > Mailing lists > Public > public-webplatform@w3.org > January 2014

Re: Error 503 Backend is unhealthy

From: Renoir Boulanger <renoir@w3.org>
Date: Thu, 16 Jan 2014 04:28:04 -0300
Cc: Etienne Lachance <el@elcweb.ca>
Message-Id: <ED72A7A2-9685-4678-B55B-FF3E81BB2EE6@w3.org>
To: List WebPlatform public <public-webplatform@w3.org>
Hi all,

Here is an heads up on the situation.

Please, do not hesitate to advise me and Etienne on this thread if you see an issue with the site.

What has been done recently;
1. Etienne worked on some OpenStack cluster limitations to help us, and fixed a disk space problem we were having
2. As Ryan suggested, I increased the error threshold and allowed connections in Fastly (caching)
3. Etienne is actively checking the infrastructure for me

But, tonight, I had to have a look and improve a few quirks in the infrastructure.

I hope it will help with the recurring errors we had in the last days.

What I did:
4. Created VM piwik2 with a larger provisioning than piwik1
5. Set piwik2 and piwik1 IPs in the zone, and propagated the change
6. Enabled project.webplatform.org Fastly config (it was not used)
7. Created project1 with larger provisioning than project

The VMs [piwik1, project] were made small  (1 vCPU, 2 Gb vRAM) due to a limitation of the resources while Etienne was building the OpenStack cluster. Since we were in holidays and the VMs were holding the load, I did not believed we should have more than that. But it seems that the load actually requires those changes.

With those changes, we might get some breathing room. 

Renoir 
~


Received on Thursday, 16 January 2014 07:28:13 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:20:57 UTC