Hi, We maintain a devpi infrastructure (one master + several globally distributed replicas) and are facing a scheduled outage of our master server due to external conditions. Devpi version info: * devpi-findlinks-2.0.0 * devpi-server-4.2.1 * devpi-web-3.1.1 We will be able to shut down our master in a controlled manner and duration of outage is expected to be less then 5 mins. Also configuration of master (ip, dns, ...) will not change. Given that I'm no very deep into devpi, I have a couple of questions/assumptions regarding impact and system behaviour, which I'd like to get your input on: While the master is down, I'd assume the following operations to fail: * devpi login to replica * devpi upload (with/without docs) to replica While master is down, I'd assume the following operations to work: * pip install -i URL_TO_REPLICA/user/index * browser access to indices/release files/docu via URL_TO_REPLICA/user/index Are these assumptions correct? After booting our master successfully, would any manual interaction be required to ensure previously (before master shutdown) operational state? Cheers, /stephan
On 4 Aug 2017, at 9:51, Obermann, Stephan wrote:
While the master is down, I'd assume the following operations to fail:
* devpi login to replica * devpi upload (with/without docs) to replica
While master is down, I'd assume the following operations to work:
* pip install -i URL_TO_REPLICA/user/index * browser access to indices/release files/docu via URL_TO_REPLICA/user/index
After booting our master successfully, would any manual interaction be required to ensure previously (before master shutdown) operational state?
Those assumptions are correct. You don't have to do anything on the replicas. They will poll until the master is available again. Regards, Florian Schulze
Hi Florian, Thanks for the confirmation and all the work you guys have put into devpi. :-) Cheers, /stephan ________________________________________ From: Florian Schulze <mail@florian-schulze.net> Sent: Friday, August 4, 2017 10:43 To: Obermann, Stephan Cc: devpi-dev@python.org Subject: Re: [devpi-dev] yet another devpi maintenance question On 4 Aug 2017, at 9:51, Obermann, Stephan wrote:
While the master is down, I'd assume the following operations to fail:
* devpi login to replica * devpi upload (with/without docs) to replica
While master is down, I'd assume the following operations to work:
* pip install -i URL_TO_REPLICA/user/index * browser access to indices/release files/docu via URL_TO_REPLICA/user/index
After booting our master successfully, would any manual interaction be required to ensure previously (before master shutdown) operational state?
Those assumptions are correct. You don't have to do anything on the replicas. They will poll until the master is available again. Regards, Florian Schulze
participants (2)
-
Florian Schulze -
Obermann, Stephan