I have an application running on top of Celery with redis as a broker. Redis is hosted at our datacenter but the workers are hosted on a separate line. We have setup two-way communcation between the servers by utilizing a vpn-tunnel.
The problem with this is that it seems that the workers are sensitive to VPN-renegotiations. Our server provider does this every hour which causes our celery-workers to be disconnected from the cluster.
Does anyone had any similar problem? Is there something I'm missing? I find it very odd that my celery-workers die because they are basically polling the redis-instance.
The VPN is setup using IPSEC site to site with a fortigate r60d.
Thanks,