6,989 reputation
21741
bio website
location Boston, MA
age
visits member for 5 years, 6 months
seen Nov 11 at 7:54

Apr
19
comment New office cabling: should we use CAT7 or CAT6a?
FWIW -- I've run 1gb on cat5 and I've run 10gb on cat5e that I made myself. That is actually one of the things that makes me really wary of these standards and the actual installs. The standards make it so a catX install will work in all situations -- reality is worse -- it will work when you "test" it with a couple nodes then act weird when you add the 80th node to the network.
Apr
14
comment Recursively delete empty folders with verbose output
I worry about the logic of this -- is a directory empty if it only contains directories that are empty? Also -- as far as logging the output -- whenever I do something like this that has the potential to wreck a whole bunch of stuff, I have it output the command it would run but not actually run the command itself. Then, if I actually want to run those commands I can just run the output like a shell script.
Apr
7
comment Anycast DNS — how do you deal with TCP DNS requests?
Which OSPF implementations offer per-flow ECMP instead of a per-packet ECMP? In a perfect world I'd just get the ECMP to chose a path based on a hash of (source IP / source port).
Apr
5
comment Something is burning in the server room; how can I quickly identify what it is?
+1 for the buddy system. I think it's a little nuts that there are DCs out there that use the EPO to also dump fire suppression as well. There are plenty of situations where you'd want to EPO without wanting to dump halotron all over the guy getting electrocuted. An EPO is a serious deal but isn't a "destroy everything in the DC kinda deal" or at least shouldn't be. The guys in the DC should hopefully understand the big red button and the fire suppression system well enough to weigh the effect of hitting the button. An EPO may actually stop a fire and save the DC, for instance.
Apr
3
comment Anycast DNS — how do you deal with TCP DNS requests?
Ah, so the connection tracking is the key -- people.netfilter.org/pablo/docs/login.pdf -- each node runs a firewall that knows the state of each other firewall and forwards traffic to the proper firewall if it happens to receive traffic "for my ip but not for me".
Apr
3
comment ZFS / ZFS send for database replication
Let's assume linux and/or freebsd.
Nov
27
comment Cisco IOS router config — how to disable SSH / SNMP on all but loopback address?
That seems really tidy. Thanks!
Nov
21
comment Cisco IOS router config — how to disable SSH / SNMP on all but loopback address?
So this means the NMS would be able to SNMP walk / ssh any of the IPs on the router?
Oct
18
comment Zabbix not getting data for one filesystem
What is in the agent log? 1.x or 2.x? Active or passive check?
Apr
19
comment How the heck is http://to./ a valid domain name?
no -- what I find odd about that is that at the time I did that lookup tt was returning someone's home computer. rr.com is roadrunner, an end-user ISP. Maybe they also offer other services, but still a bit wacky to have an MX pointing to an rr.com address.
Feb
5
comment How to debug the integration of a Windows 7 machine into a Samba 3 domain?
Why would they care if you setup your own DNS server? It isn't as if you're making anyone else use that DNS server.
Sep
8
comment Load testing wireless LANs
Thanks! I'll have a look at that
Jun
13
comment Dual Controller SAN 8x1GBE or 4x10GBE Performance
It looks like the 6224 is a line-rate switch, so that's good. I've used Nortel, Extreme, and Enterays switches -- they often use exactly the same chipset but use their own software on the switch and there are huge differences in the quality of the software. Fortunately, if you're doing pure L2 you won't need to worry about that too much. If possible, get your vendors to each loan you the equipment they want to sell you and implement it with both and load-test both. Regardless, you want 10gig if you can get it.
Jun
13
comment Dual Controller SAN 8x1GBE or 4x10GBE Performance
I think the cisco 2960 is an oversubscribed switch with a total backplane speed of 32gb/sec. You really don't want to oversubscribe your storage network. There are plenty of switch vendors out there who will sell you something that is line-rate for all ports...
Jun
11
comment Any options out there for kerberized ssh client on windows?
@alex thanks I'll take a look at it. Hopefully it has fixed the weird problem with forks of putty where accessing a non-kerberized host with GSSAPI turned on would cause a 45+ second pause before attempting any other authentication.
Jun
10
comment Collectd retrieving only NaN on Xen dom0
Do you have the same version of collectd running on the dom0 host, and probably more importantly, do you have the same "types" files on the domU and dom0 hosts? Briefly try turning on a "CSV" logger as well as the RRD logger and look at the output of the dom0's csv files. Are you just using the collection.cgi thing for looking at the RRDs?
May
10
comment Better/alternate front-end/graphs for munin?
@sciurus: First let me say that as of right now, I haven't actually finished this so I may change it around completely... But my feeling is that if I've got both the RRD and a raw CSV available, I'll have more options for a graphing back-end than if I only had one. And I suspect I'd have an easier importing / feeding a CSV into whisper than an RRD, or at worst, they'd be equally hard / easy.
May
10
comment Better/alternate front-end/graphs for munin?
I recently did "read up on status / monitoring / trending" compare and contrast and decided to use collectd for the majority of our systems monitoring. You can output CSV files and rrd at the same time; my plan is to export the CSV into something like graphite after I've got all the rest of everything worked out.
May
9
comment mirrored filesystem across a few servers
The key to understanding OpenAFS is that it is a cache management system -- there is one namespace (IE only one "file" exists) but there are cached copies of the file everywhere and a protocol to make sure that all the cached copies are consistent. If you lose the master you can turn one of the cached copies into the master, but it isn't ideal to be in that situation.
May
9
comment mirrored filesystem across a few servers
Yep. And it is possible, though annoying, to recover from losing the primary master. But it is trivial to have multiple writers and the resulting blocks stored on multiple hosts.