Clear Container - Block Storage passthrough on container
by Dimitri Bellini
Hi,
I'm new to Clear Container technology but not so new to KVM, i meet this
project because i will try to find a new to create a sort of "Isolated
environment" for a Big Bank.
They actually run multiple services on Huge Blade servers with a lot of
shared disks (more than 200 Luns per server) for every services. They
actually use a traditional "cluster" setup with resource group to
isolate/manage ip,mountpoint,services into the cluster.
In the beginning i supposed to use a container to gain the isolation and
management for this big enterprise but as soon i read and study how work
docker/lxc (and orchestrators) i found a huge "feature lack" for LEGACY
workloads.
We could not directly manage the Block Devices inside the container... and
most of the work is made from the guest host... (lvm management)
So during my research on internet a met Clear Container and immediatly i
liked the idea of a kvm engine for docker because in my mind i supposed to
"emulate" a complete system (like KVM/Qemu full version) with a very little
footprint and "finally" i can use passthrough on my SAN luns to the
container but i did not found nothing about it on the main Project Page.
Someone can give me a suggestion or maybe give me a way to solve my problem
with persistent storage?
Thanks very much
4 years
any idea how the socket channels might have gotten closed?
by Whaley, Graham
Hi,
I'm working on tracking down, or at least clarifying the 'rm stability' issue:
https://github.com/clearcontainers/tests/pull/491
It's a bit of a bug bear of a bug - different pattern of failure for each run etc.
One pattern I found was a timeout when talking to the VM, so I opened this:
https://github.com/containers/virtcontainers/issues/390
and modified my local timeouts to 10s (whilst I carry on debugging).
The next problem I've seen that I don't understand is getting one of these in my runtime log:
time="2017-10-02 15:55:42.595593471 +0100 BST" pid=23999
name="cc-runtime" level="error" source="runtime" msg="No multicast
available for CTL channel"
I had a dig, and it would look like they hyper layer has been told to close the sockets, but then somebody has tried to use one.
I don't know that area of code well, but it smells of a race or threading/locking issue - anybody who knows that code a little better
Have any ideas how we might have gotten to a situation where the CTL socket had been closed, but still notionally in use?
I'll do a few more runs to see if this turns up often now I have the timeout extensions in place - and if so, I will start to breadcrumb more debug through the relevant paths.
Thanks,
Graham
---------------------------------------------------------------------
Intel Corporation (UK) Limited
Registered No. 1134945 (England)
Registered Office: Pipers Way, Swindon SN3 1RJ
VAT No: 860 2173 47
This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.
4 years
