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.
Comments
RDNS shouldn't be buggering up but who knows, maybe a part of it didn't boot up properly. Try updating RDNS entries again? I'm thinking rc.d was silly and booted powerdns before SQL.
Francisco
I have the new series it seems, but before I was getting Network Unreachable, now another error when trying to ping / get access out :
From 2607:f358:1:fed5:39::3 icmp_seq=2 Destination unreachable: Address unreachable
Hope that helps. LV-node38
Yep, OK.
Anthony just has to update the host node and I need to hook up BGP.
Francisco
For everybody who has problems caused by the non-working ipv6, put this in your /etc/gai.conf
@Francisco let me know what you find, as we seem to be experiencing random drops in speed there and want to see if it is just us.
It could just be 1 path. I know Rob has an extra 10 gig on the way and is just pending the fiber being run.
I'll get some monitors running against the brocade soon and i'll find out exactly what's up.
It could very well have just been a stupid amount of multicast traffic. I didn't spend a lot of time dicking around with the brocade since it's just in layer2 mode. I'll know more tonight for sure
Francisco
@Francisco SSH is being extremely slow to establish a connection for me. It's taking an average of 18 seconds. Once it's connected, there are no delays at all.
'usedns no' In your sshd conf file and restart SSH.
If that helps then your connecting ISP has some slow RDNS servers. It's also possible I need to tune named some more
Let me know.
Yes nice and fast again :-)
It was only after the move that I noticed any slowness though, perhaps a coincidence.
Unlikely Please, log me a ticket 'Slow DNS resolvers' as the title and link this comment I'll check into it once I'm home. I'm fairly sure it's just BIND wanting moar RAM. It has 500MB but I'm thinking that's simply not enough.
Francisco
Fran, network is still really slow. @lbft confirmed that as well. Node 19, do you want an IP to be pmed?
It was really quite slow earlier today (something like 1 MB/s over HTTP from my VPS on storage01 to my Versaweb dedi in the same DC) but just now when @Wintereise mentioned it I tried again and it seems to have picked up, hitting 13 MB/s.
This is quite ironic. @Francisco moved out of SJ because of network performance, and we're experiencing more issues in LAS.
I'm getting ~1.2MB/s download speeds from my VPS(node 55).
Well it could be a few things
And sure, PM me that wint.
I've been seeing a lot of ugly things I don't like so I'm seeing if I can clean it out of the network.
We've been seeing more and more attempts at dns reflection attacks going on so I'm wondering if it's causing enough PPS to slow some things down.
I'm working on getting SNORT setup right now to clean it out.
It could also be a configuration issue on my primary ethernet port. I've already emailed Rob asking him to confirm his MTU settings.
I may just rush the 10Gbit port facing FH so then I can just run jumbo frames on both ends.
Francisco
I pmed ya so you can make sure, but things seem to be getting better slowly :x
I'm not changing anything, i'm chipping away at my TODO list first.
I'm just picking out a 2nd 10gig card right now. I should have snagged it when I was in LAS, ah well.
Francisco
No network problems here? lv-node-29.
I don't think it's so much the nodes as just busy times for our ports. LAS' network is just just faster so we're getting to a point we were waiting to see. I'll hear from Rob come morning about where we can go from here. I got a few ideas already going on.
Working on v6 right now.
Francisco
@mojeda Pipe the output to /dev/null
/lazy
down again
Nothings down over here?
I've been working for hours now.
I know a node or two threw up but that's about it.
Francisco
looks all good from http://buyvmstatus.com/
Just that one lv kvm node seems to go up and down, possibly a non issue
Does FiberHub even have 10Gb?
They have like 60Gbit+ <_< Abovenet is due some time in the next month or so, so another 10gbit of that.
I'm just waiting for Rob to let me know what connector he needs from me. I'm guessing it's just SPF+.
When we were in there in Oct they had at least 40 gig in use.
They got some huge streaming guy there that accounts for a lot of it and he's getting ready to ramp up a crap load. HE also has their POP for Vegas in there in their MMR.
EDIT - Fixed the month we were in there previously
Francisco
V6 has been brought online in both LAS & NY.
I just finished running a mass update script updating all stallion IP entries so they should be correct now. I'll likely run a mass update script copying over all RDNS entries in a little bit.
Francisco
Very doubtful, but anyway, I think it was a bad move on your part as the problems you have been seeing will not go away. FiberHub alone only has 22,000 ip's, pretty much the same size as myself.
Small time DC, small time commits = problems
EGI only had like 20Gbit and of that only a % was for the rest of their clients. They have at least 1 large time client that uses like a 10gig or something close to that.
Speeds are better and the limits we're hitting now are because we need to upgrade our own port, not something on the DC's end
We got more rack space (a cage now), our new brocade is online and a few extra nodes are coming up for sale.
I wasn't happy to have to do the move but I'm happy with the end results
Francisco
If I recall right, the brocade you got doesn't even support 10Gb connectivity, and as for the cage I'm sure they were dying for someone to take the space.
The mix of BW is worse then EGI too.
It has 2 10gig ports on it <_<
Sure, if we got egi's full blend. Fact of the matter, though, is that EGI was forcing our routes to only go over HE. Whenever HE had issues we felt it heavily. We had VERY minimal nlayer routes and almost no GBLX routes.
My usual example is this:
Egi always said they would 'look into it', but the fact of the matter is that the only reason I'd ever take HE to get to nlayer is because they were forcing my routes to go that way. BGP is strictly about shortest path and going direct nlayer would have gone that
Francisco