New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
BuyVM Down?
Just noticed BuyVM is down again in LV
Comments
Can you read? Status site clearly says maintenance.
Network/Router Maintenance. - 10/07/2013 21:00 UTC (started 43m ago)
Well, that link doesn't load for me, because their website is hosted in LV I assume. I'm interested in hearing what that is about if anyone knows.
Internet tip: prepend a url with cache: in google.
Eg: cache:https://my.frantech.ca/announcements.php?id=159
You should follow their Twitter. They're adding a 10Gbit uplink to their router and removing their LACP setup.
Thanks @vld I tried that before creating this thread but it took forever to load, after waiting about a minute it loaded. It was only planned for 10 minutes but has been down for longer than that it seems. Good to see that upgrades are happening tho.
Servers are back up now.
Were servers rebooted due to this network maintenance? Seems like the uptime on BuyVMStatus reset, unsure how that works and if it goes by the node uptime or not.
No, the announcement said only the network was going down. I believe uptime on buyvmstatus is based on the last successful ping
From the footer: "This site is not operated by, sponsored by or affiliated with BuyVM in any way. Information may be inaccurate."
I'm aware, thanks though, its appreciated. :-)
I was just curious how the monitoring worked, and if the servers were rebooted or not.
All 3 of min were not rebooted.
I got 3 on LV
Checked uptime, they wasn't rebooted
Just network taken down for upgrade
There was no reboots on that.
The router maint went quite well A few derps since I forgot some of the quirks in Quagga/Vyatta though. I had a single /24 online but couldn't for the life of me get the bigger announcements passing until it finally clicked.
Oh well, LACP is long dead and the new router is operating very well. We have to patch around Cogent in a few areas but past that I'm extremely happy with how this router is handling.
Francisco
@Francisco let us know how much MORE traffic that ends up pushing later. Interested to know.
Haha, you sneaky one, want to find the deficit it had till now... :P
Nah, Fran quoted me some numbers from the their previous configuration. I just want to know how much it improved in comparison.
Chairman Mao, give me a break
I don't expect it to jump to us pushing 4Gbit/sec or something. If you have 4 people downloading a 4GB file within an hour and all 4 go at 100K/sec, you'll end up with a much higher 'constant' than if they burst a bit to get up to multi MB/sec.
I know personally I've jumped to 4MB/sec from 400K/sec.
Not 'till now'. When we moved from EGI we pushed about 700Mbit/sec and rarely go anywhere near even 1Gbit/sec. Short of throwing a flood out the door there was no way to push it 'past' that point.
Once in LV we got the router racked up and peaked around 1.3 - 1.4Gbit/sec of usage with LACP's. At that point we were simply suffering from the hashing the CPU has to do as well as the terrible interrupt handling the e1000 nic's have. We bought a nice brocade we were going to use but at only 2 x 10Gbit ports we ran out of ports when trying to do port mirroring. The brocade would have done LACP w/o issue i'm sure since its hashing is handled in hardware, not dumped on the CPU.
The new box is a dual hex core E5 2620 w/ Intel x520's. We got 4 x 10gbit ports and 6 x 1Gbit ports using IGB* based drivers which have proper queue support. We sit at 0.0 - 0.2 loads at 1.2 - 1.4Gbit/sec. Autonull munches on some CPU so we're now sitting at < 0.5 loads (half a core in other words).
Francisco
@Fransico which NIC are you using? X520-LR1 from Intel?
We have 2 x X520 dual socket cards in there and then 2 350T's in there.
The onboard also uses IGBE but we don't have any of the 1gbit ports in use as they're not needed.
Francisco
I always loved software routers, never used anything else except for home connection where some OpenWRT/DD-WRT/Tomato does a good job without much hassle.
In the good old days the router was also doing layer 7 QoS and filtering and 100 mbps line was really heavy, load was jumping to 4-5 for a single threaded (Duron 1600) machine and latency accordingly
Things changed, now cpus are much more powerful
Hashing is where things get iffy in software. Our NY router sits at like 0.3 loads and is using E1k's. The difference is since there's no hashing so we're able to improve the E1k's interrupting Intel's drivers handle interrupting nicely but when LACP is in place you can't take advantage of it.
We had to move to freebsd just to get the e1k's to not crap themselves during a 300k pps SYN flood. Back in Awknet we had a vyatta box and once it hit 150k pps we had ~30% packetloss; brutal stuff for sure.
Francisco