[pve-devel] [PATCH manager] fix surviving update store

Emmanuel Kasper e.kasper at proxmox.com
Tue Jul 5 15:05:40 CEST 2016


On 07/04/2016 01:16 PM, Dominik Csapak wrote:
> when a load of an update store was ongoing and stopUpdate
> was called, the task could not be canceled, and the store
> would remain indefinitely

I can reproduce the problem you mention here  ( when a node in a cluster
becomes unavailable in the network, background UpdateStore ajax requests
to get the statistics time out after a long delay and block all other
pending UpdateStore ) and the patch fixes indeed the problem.

What I am wondering is why do we treat the UpdateStore requests serially
in the first place ?



More information about the pve-devel mailing list