#arpnetworks 2019-04-18,Thu

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
***up_the_irons has quit IRC (Ping timeout: 268 seconds) [01:15]
.......... (idle for 46mn)
kleszcz has joined #arpnetworks [02:01]
........................................................................................... (idle for 7h33mn)
ziyourenxiang__ has quit IRC (Ping timeout: 246 seconds) [09:34]
.......................................................... (idle for 4h49mn)
tmarble has joined #arpnetworks [14:23]
tmarbleI can't get into the portal... can anyone help? [14:23]
........ (idle for 36mn)
***up_the_irons has joined #arpnetworks
ChanServ sets mode: +o up_the_irons
[14:59]
up_the_ironswonder why I was booted... [14:59]
tmarbleup_the_irons: can you help me get back into the portal? [15:00]
up_the_ironstmarble: just sent you a response to your ticket
tmarble: in fact, you should have a password reset email in your Inbox
I generated one just now
[15:01]
tmarbleyep, trying it
yea, i'm in!
[15:02]
up_the_irons: thx! (wondering how you can do the live VM migrations??) [15:11]
mercutiolive migrations mean copying memory contents over network
so basically it just looks at the state of your vm, and recreates that on another host
the storage itself can be accessed from multiple locations so doesn't need to be replicated
[15:19]
tmarblethat's crazy! is it with lxc, virtbox, xen or something else? [15:20]
mercutiokvm/qemu [15:20]
tmarblenice! [15:20]
mercutiobackend is ceph [15:21]
tmarblei.e. you kind run "swap" to ceph and let it do the "network distribution" for you?
.. but page out everything
[15:21]
mercutioceph stores the whole block device
and distributes it across multiple servers
[15:21]
tmarbleoh.. the disk sure [15:21]
mercutioso you can lose one server and keep going etc [15:21]
tmarblehow about memory? [15:21]
mercutiobut it also means that you can join the network between two locations
and access the same storage from another location
and storage can be moved from one location to another with no down time too
memory it basically justs read in all of the memory and sends it to the other host
[15:22]
tmarbleso literally there's only a couple of seconds (to risk of transactions in flight) to pivot the "master"? [15:23]
mercutiothat can take a few seconds, so after it's done that it copies any new changes [15:23]
tmarbleis that something qemu supports out-of-the-box? [15:24]
mercutiowell it's like swap, it'll trap all new memory accesses
so it knows to send them
umm sort of
it's out of the box.. but ceph, etc is reasonably complicated to setup
[15:24]
tmarbleunderstood [15:25]
mercutioyou need at least 3 or 4 storage hosts to run ceph storage
but it scales out nicely
[15:25]
tmarbleit's the free/open source equiv of S3 :) [15:26]
mercutiobut it means that very small setups aren't really used with it
ceph has a few things, an object store like s3 is one of them
it also has rbd which we're using, and a file system
rbd is basically optimised for block device storage
and supports useful things that cater to that like resizing the block device
[15:26]
tmarblecool! [15:27]
....... (idle for 33mn)
***ziyourenxiang__ has joined #arpnetworks [16:00]
................ (idle for 1h18mn)
mnathaniWhats the network capacity / link between the 2 cages? And is it routed via the Meet Me Room, or direct cage to cage? [17:18]
................................. (idle for 2h44mn)
***hive-mind has quit IRC (Ping timeout: 268 seconds) [20:02]
hive-mind has joined #arpnetworks [20:07]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)