46,889 reputation
25147293
bio website kbrandt.com
location Hollywood, FL
age 32
visits member for 5 years, 2 months
seen 7 mins ago

After about a year of being a member of the community I am now Stack Overflow Valued Associate #9 (Or Server Fault Valued Associate #1 if you like).

Besides trying to keep these sites up I will also be writing posts for the [Server Fault Blog][2].

[2]: http://blog.serverfault.com/

@KyleMBrandt


Jul
2
comment Windows TCP Window Scaling Hitting plateau too early
At the rate we are all going it looks like the correct answer is going to be "Switch all your clients to Linux" ;-)
Jul
2
comment Windows TCP Window Scaling Hitting plateau too early
@Pat: You can click the vote number too see the breakdown of Upvotes/Downvotes. Currently you have no downvotes on your answer. My answer doesn't solve his problem (but no answer does yet), it does explain and localize the problem (hopefully correctly!) which is an important step in troubleshooting.
Jun
30
comment Windows TCP Window Scaling Hitting plateau too early
@SmallClanger: Another guess, try playing with netsh int tcp show heuristics / netsh interface tcp show heuristics disabled|enabled
Jun
30
comment Windows TCP Window Scaling Hitting plateau too early
@SmallClanger: Oh, and to help clarify, my graph if I understand it correctly displays the RWIN as it currently is over time. In other words, it is "currently available bytes" in the buffer. So the RWIN always has plenty of space.
Jun
30
comment Windows TCP Window Scaling Hitting plateau too early
@SmallClanger: Also look at "Expert Analysis" in the two captures. In the Linux-to-Linux one, you see Zero window messages, duplicate acks etc. This is actually what one would hope to see as TCP gets pushed to its limit and then enforces some flow control. In the Windows to Linux case, everything is clean because TCP isn't "stress tested".
Jun
30
comment Windows TCP Window Scaling Hitting plateau too early
@SmallClanger: Regarding "Could there be some other reason that Linux is artificially limiting the RWIN?". I'm not positive, but I would agree with your theory that the RWIN doesn't increases because it is never getting low. It would be silly to do so, because that would mean more memory usage when there is no reason to take more memory for the receive buffer.
Jun
29
comment Windows TCP Window Scaling Hitting plateau too early
Any chance of putting up some packet captures from one of the slower senders somewhere?
Jun
19
comment centos + php leak memory, but where?
@AndreaSica: Re MySQL. I'm not positive, but I don't think MySQL would be cache, I would think it memmaps the the database files so it has more control. Higher level languages like PHP will be allocating memory with things like malloc(), they just do it for you. How about a screenshot of top + free -m output for each the following: 1) After a reboot and before your script has run 2) While your script is running, and 3) After your script has completed
Jun
19
comment dump tcp connections without tcpdump
@Isaac: Also netstat, if they won't let you have netstat you might want to start searching the job market before you go nuts ;-)
Jun
18
comment How does vm.overcommit_memory work?
@MikhailMorfikov: Yes, practically I think 0 makes the most sense. In my environment, the only time I enable 1 is for Redis, which does stuff where it expects do be asking for a lot more memory that it is using due to a fork(). The child will pretty much use all the same memory pages, but Linux doesn't know that says to be safe it has to assume 2x memory will be used (if you want to learn more: redis.io/topics/faq)
Jun
18
comment How does vm.overcommit_memory work?
@MikhailMorfikov: In theory, I believe so, but who knows what these programs are doing. Would want to see a more controlled environment with a simple program that just allocates say a gig of ram via Malloc. And then run the test after rebooting between tests.
Jun
16
comment What is the maximum theoretical throughput of a NAS device to 100 workstations with 1Gb/sec connections?
Ya, just keep what TomTom said in mind, VDI is likely highballing it. But when you lack information, and you have the budget, better too much than too less.
Jun
16
comment iDRAC login loop
Also, if there are too many log in attempts with a false password, iDRACs will block people from logging without telling them for what seems to be an undocumented amount of time....
Jun
16
comment What is the maximum theoretical throughput of a NAS device to 100 workstations with 1Gb/sec connections?
You could also assume their entire workstation is hosted on the SAN, and pretend you are building a Virtual Desktop Infrastructure (VDI) and use some estimators like blogs.citrix.com/2010/10/31/… .
Jun
15
comment Why is Router IP Address Repeated in Trace Command?
@FelixFrank: In addition to the routing table, capture captures with a tool like tcp might provide insight. In particular observe the IP header.
Jun
13
comment Agentless monitoring: how does it work? Advantages over traditional monitoring?
@syneticon-dj: Eventually :-P
Jun
11
comment Find out ESXi host from VM?
@mfinni: Because we have monitoring agents on every Windows / Linux host pushing to the monitoring system, and currently don't have anything using the vSphere API as of yet. So this is likely more convient as thing currently stand.
Jun
11
comment Find out ESXi host from VM?
This is for monitoring. I'd like to tag every VM with its Current VM Host. So I want an option where this is updated at runtime and not hardcoded into the VMs config. It doesn't have to be instance, could be something refreshed every X minutes
Jun
7
comment Choosing the right AWG for Server Power Cables?
I don't like them myself, but I'm usually not the one working in the datacenters, so I leave it up those guys.
Jun
7
comment Choosing the right AWG for Server Power Cables?
We used 14s, combined with cable arms, that is a bad choice. They end up popping out of the servers even when you velcro them in nice and tight. We are redoing everything with 16 now (Which is thinner). So this is something to keep in mind.