356 reputation
311
bio website
location Houston, TX
age 31
visits member for 3 years, 6 months
seen Oct 24 at 20:25

Systems Engineer & Software Developer


Aug
28
comment Can TCP and UDP packets be split into pieces?
Might be worth clarifying the last bit for novice readers: you will see the complete data for the datagram in question. If any one of the split packets gets lost, the datagram is lost and the UDP layer will never know of it. As long as all packets in the datagram are received, they will be assembled at the IP layer and then passed up to the UDP layer. This doesn't preclude the possibility of missing "chunks" in the datastream. Not to be a pedant, but when I was learning this stuff I didn't grok the difference between IP frag and UDP loss until the 2nd or 3rd pass through the textbook.
Mar
28
comment Why does a hardware router perform better than a Linux router with better specs (RAM & CPU)?
+1. Add to that the fact that Cisco has the engineering resources and know-how to replace software with hardware when needed. Meaning, if a particular operation is slower than they'd like in software, they can put effort into engineering an ASIC, adding instructions to their existing processing units, or build a hardware module to help speed it up. The tables tip overwhelmingly in favor of performance when you have the luxury of controlling both the software and the hardware.
Dec
17
comment Forwarding udp ports iptables packets “lost”?
@SeanC: iptables does have logic to allow it to track UDP conversations‌​.
Nov
27
comment CentOs vhost shows default apache page
@Bernhard: See my edit. You have a permissions problem, run the command at the end of my answer to fix it.
Nov
27
comment CentOs vhost shows default apache page
@John: Answer edited, though I have to play devil's advocate here and ask: Why would anyone want to re-enable SELinux? :)
Oct
24
comment What's the best way to check if a volume is mounted in a Bash script?
Possibly worth mentioning: if you use /proc/mounts at a lower level than a shell script (say, parsing with Perl or Python), you must remember that /proc/mounts is only atomic within a single call to read() -- so be sure to read the entire file and then parse it, rather than relying on parsing line-by-line from the file handle itself.
Oct
3
comment Should I reinstall my production server from 32 bit to 64 bit if it has 16GB of RAM?
@Olipro: My apologies, my intent was not to offend you or denigrate your answer. I just tend to prefer correctness over conciseness so felt the need to clarify your comment about the instruction sets. At the end of the day, switching to a 64-bit kernel and using the 64-bit binaries of important packages will certainly benefit the OP in multiple ways and should be done except in specific scenarios (none of which seem to be the case here).
Oct
3
comment Should I reinstall my production server from 32 bit to 64 bit if it has 16GB of RAM?
@Olipro: Yes, I'm aware of the SSE2 guarantee on x86_64, but your answer as worded suggests that x86_64 is faster than x86 solely because x86 is an "old" technology. My comment was intended to clarify that point. Also, switching from 32 to 64-bit OS on the same x86_64 system will have little noticeable impact since the CPU extensions available in 32-bit mode are there regardless of the OS. But, changing CPU from an IA32 to an x86_64 may provide a benefit, assuming that the IA32 CPU doesn't provide the extensions that the x86_64 does, and that x86 code is recompiled for the new extns.
Oct
2
comment Should I reinstall my production server from 32 bit to 64 bit if it has 16GB of RAM?
"...generic 64-bit instruction set has generally better performance than generic 32-bit (due to the sheer age of x86)" -- This is a dubious statement. While it's true that 64-bit code is generally faster than 32-bit code (for various reasons including register size, memory bandwidth, and better optimization), the performance difference has nothing to do with how long x86 has been around. In fact, x86_64 is an extension of x86 and it therefore could be argued that the instruction set's history stretches all the way back to 1986. :-)
Aug
27
comment I've inherited a rat's nest of cabling. What now?
@NReilingh: I was being a bit flippant, the wording of rnxrx's comment makes it seem like he was suggesting using "premade certified punchdowns", which is obviously impossible with current technology.
Aug
27
comment I've inherited a rat's nest of cabling. What now?
@rnxrx: Where do you get certified, premade pre-punched patch panels? (yay, alliteration!) I would love to be able to drop in some working panels with the cables already run, rather than pulling all that pesky solid core cable through the walls.
Aug
22
comment Successfully, permanently ban a user from a website?
Good suggestions here, but the point about "outing" the proxy by use of a nonstandard port request is probably well within the realm of "hurting legitimate users more than hindering malicious ones". That said, for a social networking website it may not make a huge difference as the majority of your users are probably not behind a firewall.
Aug
15
comment Does a RAID 1 really protect from HDD failure?
+1. See also: serverfault.com/questions/2888/why-is-raid-not-a-backup
Jul
16
comment Why does heavy I/O slow my server to a crawl?
My initial results with switching to deadline are curious indeed. I saw mostly the same behavior up to a point: load average went to 9, and the array and layers above showed vastly more tps than the disks themselves. But, after a few minutes, the load average dropped back down to around 2.1 and the system was responsive again, buffer memory was still around 1.5 GB and dropping slowly. I'm going to test this multiple times but it seems that the mkfs is filling up the buffer and then bogging the system down with I/O wait until the writes to disk catch up.
Jul
15
comment Why does heavy I/O slow my server to a crawl?
@AndrewSmith: I've dried on different hardware, both an IBM SystemX 3250, Dell PowerEdge 2950 and Dell PowerEdge R410, with both SAS and SATA drives of differing speeds. The IBM and the 2950 have non-battery-backed LSI SAS controllers, and the R410 has a PERC6. NCQ is off in all cases (we have to turn this off due to an unrelated issue). I honestly don't believe this to be a hardware issue.
Jul
15
comment Why does heavy I/O slow my server to a crawl?
@MichaelHampton: No, there are two disks (/dev/sda1 and /dev/sdb1). The above is supposed to be a visualization of the full stack. I'll edit the question to clarify this.
Jun
1
comment What is a suitable simple, open web server for Windows?
Suggest changing question title to: "How can I overcome my irrational fear of Apache?" That said, I've been running Apache on all of my systems (Win, Lin, OSX, Solaris) for well over five years doing a variety of different things. My desktop runs a development instance with modules galore and a simple instance serving binary files on a configurable high port to my HTPC. (sound familiar?) It will get the job done with minimal fuss, and if you need it to do more in the future there's no need to migrate to a "more capable" server since it's already there.
Mar
21
comment CentOS expanding directory
I'm surprised no one has mentioned using LVM.
Mar
21
comment CentOS expanding directory
I'm not sure if this is evil or genius. Evil genius?
Jan
6
comment To fsck or not fsck after 180 days
"Switching to xfs would at least allow for a reasonably fast fsck"... Did I miss something?