That I really wish Google was willing to test outside of the US for it's 1 gigabit per second, fiber-to-the-home connections
what did you learn today?
- Thread starter chalex
- Start date
- Status
Originally posted by jaericho:
Why? It's the quickest way to find out if the change will work (or not).
Sure, as long as the General Manager isn't affected by it...
I think you misunderstand me. For sure the internal temps were at 140, because ambient was well over 100 at the time.
I just think the exhaust temp was more like 180 or so in reality. I don't think they bother to calibrate up to 280. I would whatever is the advertised operational range of the equipment would be covered and maybe a bit beyond that.
I don't think it's reasonable for a company to spend the resources to check for calibration, say twice the operating range in temperature.
And the servers didn't shut off.
The one that autorebooted was an Exchange server. Our ONLY backend Exchange server. Wooo.
Working with legacy code is such crap. Just the simple process of converting shell to perl is difficult enough, do you really need to add triple-nested if/elsif loops that do not actually do anything that I spend hours staring at because I am convinced that it had to do something? Unfortunately, elder programmer, you can no longer be fired since you quit years ago. Clearly an asshole who had it out for me
I'm now glad that our thermal shutdown settings for our HP equipment actually works (or wasn't it set up?). We're putting in another 20 ton unit because the 2 we have can't cope with the heat from the servers if one the compressors shuts down for any reason. It's not bad in the winter because we can just open the doors (or shut the heat off in the adjoining office).
No, you see, the contractor setting all that up, didn't actually set it up. It's the same thing with the UPS monitoring gear, even though they have officially "finished" with that part as of a few months ago.
That's not been set up either.
And so forth and so on.
In that case, I just learned that intentionally killing someone may be justified...
That contractor needs to have his business drug through the mud on every forum and BBB site you can find. I don't think suing the hell out of them for breach of contract would be unwarranted either.
today I learned the 4.x version of NOD32 blows. and I'm in the process of switching everyone over to 3.x (we use 2.x on the servers).
Sticking with the subject of horrible legacy code to support...
The next version of our platform runs Red Hat EL. That leaves me supporting Linux 2.6.x kernels, Solaris 10, Solaris 8, and ... HP-UX 11.0. Fuck me!
The next version of our platform runs Red Hat EL. That leaves me supporting Linux 2.6.x kernels, Solaris 10, Solaris 8, and ... HP-UX 11.0. Fuck me!
Continued healthy operation of the International Space Station depends in part on a single Dell GX1 in the SCTF building. It's a Pentium II computer with 16 megabytes of RAM, running Windows 95 and Visual Basic 5, which is used to maintain an application that serves as the interface to multiple important flight systems. It is certified for use only in its exact configuration and can never be modified.
You can't make this shit up.
You can't make this shit up.
Hah doesn't surprise me in the least. The three flight computers on ther orbiters still use Core Memory as far as I know....
I really hate working for a vendor sometimes....
Oh how I miss the days of being a normal consultant.
Oh how I miss the days of being a normal consultant.
Originally posted by ronelson:
Working with legacy code is such crap. Just the simple process of converting shell to perl is difficult enough, do you really need to add triple-nested if/elsif loops that do not actually do anything that I spend hours staring at because I am convinced that it had to do something? Unfortunately, elder programmer, you can no longer be fired since you quit years ago. Clearly an asshole who had it out for me
Code deleted is code debugged.
At the end of the day, I have to have a replacement for it. If I delete it, I have to write replacement code. If I want to write replacement code, I have to understand what the hell is going on in there.
It could be worse. Another guy I was talking to yesterday inherited a program with vars like the hash "%h" and the function "m". Lord only knows what the fuck is going on there.
I had the exact same thing happen when pulling data off of a client's "dedicated server" at a reasonably litigious Texas-based hosting company. I logged in via RDP to pushed a bunch of content from their websites onto my servers. I had everything copied over and noticed that it had never had a single update applied in years. The machine was running W2K. It needed to stay up longer but some issue popped up and it made sense to reboot. I rebooted remotely and waited. And waited. And waited.
It never came back.
I called the litigious hosting company (whose support department answers with a more generic name) and finally got to a real person. They called someone else to go run and look at the machine.
When they arrived, the machine was hosed. The drive wasn't spinning up, and they couldn't get anything to come up on the monitor they attached. We changed out the power supply and later some other parts. Even moved the drive into a different machine. Whatever went wrong ate the drive.
Talk about just-in-time backup! (they had never backed up this machine, thinking that the hosting company did that)
They did lose stored email, but otherwise, everything important had been moved to my machines.
Talk about an excellent way for the script kiddies to hide stuff in "run" Put a CRLF in the entry's name.
....in order to get an ancient app developed for VMS, now running on a modern OpenVMS install on a HP Superdome, to print successfully to a modern nonHP color laser printer:
-be sure to make the users select B&W only for printing
Sounds simple, right? That involves educating our end users. This won't go well.
If the mopier vendor had a B&W only driver, that would be better. That would force PCL 5 emulation instead of PCL5c (color extensions added, prior to PCL6, AFAICT) without requiring our end users to click on too many buttons.
This could have been avoided if the District office had budgeted for the PS printing addon for the server, or had deployed it or had told us it existed. etc.
During the building and equipment planning phase, they probably ignored this particular use case, even though the majority of mission critical data is on these legacy systems.
Instead, we're due for a Naruto style 1000 years of pain.
-be sure to make the users select B&W only for printing
Sounds simple, right? That involves educating our end users. This won't go well.
If the mopier vendor had a B&W only driver, that would be better. That would force PCL 5 emulation instead of PCL5c (color extensions added, prior to PCL6, AFAICT) without requiring our end users to click on too many buttons.
This could have been avoided if the District office had budgeted for the PS printing addon for the server, or had deployed it or had told us it existed. etc.
During the building and equipment planning phase, they probably ignored this particular use case, even though the majority of mission critical data is on these legacy systems.
Instead, we're due for a Naruto style 1000 years of pain.
In addition to possibly using the LJ4 PCL driver if you are using a print server just set the client default to B&W. We did this for cost saving reasons (~$150K per year saved).
Actually, no. It will display the whole string without the CR+LF. I have, however, no idea what happens if you use a CR+LF in a "run" entry. Might be worthwhile to try
I don't think they'll go for it. The manager level people want their pretty colors the rest of the time without having to learn anything new.
Seriously.
I think maybe the best choice is to set up a separate print queue to the same printer, but with different defaults. That will work better than asking them to change printer settings in a deeper dialogue box.
The people do seem to understand the concept of choosing different "printers" (better not start trying to explain printqueues).
At one of my previous jobs we had a guy who was a bit of everything, admin, programmer, network guy, etc etc. He was one of those people who just seem to know everything. He was a bit odd though, in quite a few ways.
For some reason he was once tasked with writing a client application to interface with our core products, I guess the "real" programmers had forgotten how to code in C or something. He wrote the code in no time at all and worked wonders, still did when I quit.
When he quit, another guy became the owner of said code, had a look at it, and damn near started crying after seeing all the variables were simply named a, b, c..., and if the letters ran out, a1, b1, and so forth.
I have no idea what functions and such were called, but I assume those weren't any better.
What sort of issues did you run into? I've been using 4.x on my clients without problems. If memory serves a few people on the NOD32 forums did end up switching their servers to 3.x though.
Also, Microsoft has published an article here which may be of some assistance.
Ouch! I think that is about where this guy is. The whole toolset has a fancy acronym and when they load a new server with it, they dump a tarfile on there, expand it in a directory...and hope it works. They are about to move from Solaris to Linux as well, hopef they are prepared. If nothing else, the path-to-shell information in each script is going to be different...or they could do something crazy like installing older versions of tools in those directories. I would not bet against that...
Are you saying 110 spindles are only capable ofr 4k IOPS? That's pathetic, but I think I'm reading your post wrong.
I think thats the point....
They aren't driving enough IO to the disks.
I just went through this with a customer who was complaining they only saw 30MB\s to the controllers. The where sure they were an DW type workload. When I gently explained that 30MB\s of 4k I\Os was a lot of activity I still got the response of my IPOD is faster than that.
Finally the application vendor was involved and they said your as hoc query of a weeks amount of data should take a long time to return results they finally believed me it wasn't an array issue.
They aren't driving enough IO to the disks.
I just went through this with a customer who was complaining they only saw 30MB\s to the controllers. The where sure they were an DW type workload. When I gently explained that 30MB\s of 4k I\Os was a lot of activity I still got the response of my IPOD is faster than that.
Finally the application vendor was involved and they said your as hoc query of a weeks amount of data should take a long time to return results they finally believed me it wasn't an array issue.
I completely agree with this.
Does it make it silly that people have "PrinterA Black and White Double-sided", "PrinterA Black and White Single-Sided" and "PrinterA Colour Single-sided"? Yes, but it works a hell of a lot better than the alternative.
Originally posted by scorp508:
Are you saying 110 spindles are only capable ofr 4k IOPS? That's pathetic, but I think I'm reading your post wrong.
Currently yes, we have a p570 with all the processors enabled, dual 2 gig FC connections and 110 or so 10/15k FC spindles dedicated to it maxing at 4.4k IOPS and they are blaming it on the array. Never mind that the Oracle binaries, log files, backup dump locations and database files themselves are all mixed together on the same file systems. It's clearly been bottlenecked by something in software for years but they insist it's disk bound.
On the good side I was able to benchmark our new data warehouse box at ~4100 IOPS with MSSQL 2008. That's not bad since it's only got 28 drives in it (RAID DP). Small block mixed writes to a 16gig test db. I am very pleased.
Yeah people some how don't understand there is a difference between streaming 1 file to 1 person and looking up and replying to multiple people at the same time from a massive number of possible locations. Including a lot of admins.
If our linux dude talks about throughput one more time I'm going to choke him. He was bitching because our main SVN box was "only" peaking at 80 mB/s so it was slow. Nevermind that the script he wrote to "clean stuff up" was pegging the CPU so badly even a ping had 13 ms of latency (on the same switched subnet).
Learned that iLo on blades controls the blade chassis fan speed and that it's a known issue with 1.80 version of iLo firmware to make the blade chassis fans go at full bore all the time. Fix was to down rev to 1.78 for every blade in the chassis. For fun...not.
1.81 doesn't fix it? I've been updating our DL360 G6 and DL380 G6 servers to 1.81. Maybe it isn't out for the blades yet.
When dealing with a dell tech over chat sometimes specifically stating the problem, the solution and what the tech should do next is not enough. This latest endeavor with Dell chat leads me to a simple conclusion: Having the dell tech remote my machine and then prove to himself that I was correct the entire time is one of the best feelings. Vindication is sinfully sweet.
I also learned that the Dell repair depot will (at their whim) change your wlan card model.
I also learned that the Dell repair depot will (at their whim) change your wlan card model.
- Status