An article on the idea behind dtrace and what exactly can it do in the upcoming Solaris 10 releases.
DTrace — Most Exposing Solaris Tool Ever
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
http://www.sun.com/bigadmin/content/dtrace/d10_latest.pdf
Both of links are very insteresting to read and it’s written in D Programming Language.
sun has always been good at innovation and quality engineering. this is why you pay for solaris.
Tons of comments against Sun because “if Solaris was any good, they’d be promoting Solaris, not knocking RedHat”.
This article about something Linux will not get for years (based on the fact that Linux has still not got a half-decent equivalent to kdb, let alone savecore) gets two comments.
Tons of comments against Sun because “if Solaris was any good, they’d be promoting Solaris, not knocking RedHat”.
This article about something Linux will not get for years (based on the fact that Linux has still not got a half-decent equivalent to kdb, let alone savecore) gets two comments.
Unfortunately, that’s because the majority of the people posting on this site are very biased towards Linux and GNU, and the moderators get sick of the inevitable flamefest’s that result from it, as do other legitimate posters.
I myself have pointed out these particular flaws of Linux here many times, pissing into the wind it was, as it’s easier for the moderators to take me to task than it would be to do so with the vast majority of their sites’ more rabbid Linux using posters, regardless of the fact that they’re the software equivelents of of those middle eastern nut jobs that are beheading people for similarly stupid religious reasons.
Yes, Solaris is great for all of the things it has that Linux won’t have for half a decade to come, but man, don’t dare say it out loud here, because the zealots will eat you alive.
Shit man, anybody know of some Sun jobs.
I’d buy Solaris 10 just for DTrace.
A very well written PDF.
This is Sun’s strength. Very high tech — very well written.
I can’t understand why anyone would run Windows if they even glanced at Sun.
I can’t understand why anyone would run Windows if they even glanced at Sun.
Game support?
I’m a Linux fan myself – used it as my desktop for 6 years now. Author of http://speedtouchconf.sourceforge.net/, for example.
But I do know Linux’s shortfalls, because I work with Solaris every day.
That doesn’t make Linux a Bad Thing – it’s a Great Thing. For *my* purposes, the lack of savecore et al aren’t a huge problem – if Linux had something approaching Dtrace, it would pretty much be a world-beater… that level of driver and app support, with a dev tool like Dtrace.
For that kind of developer support, Solaris still beats the pants off Linux. I’m not gloating about that – I’d rather that every OS had the best features of all; if Windows had *nix stability and logging facilities; if Solaris had Linux’s driver support, if Linux had reporting facilites comparable to modern Unix… we’d all be laughing.
That has to beat religious wars.
But, it’s religious wars that keep sites like OSnews and /. alive….
Emacs? Dirty piece of memory-hogging scum! Give me vi[m]!
I can’t legitimately argue with anything you’ve said there, except for
Emacs? Dirty piece of memory-hogging scum! Give me vi[m]!
you dirty vi[m] using bastard! (no, I’m not a dirty Emacs using bastard, I’m a dirty ee using bastard ;^)
If you’re interesting in playing with the power of dtrace, also check out http://users.tpg.com.au/adsln4yb/dtrace.html – DTrace Tools . I’ve confirmed their working on SPARC’s and with a 40% increase in performance by just using Solaris 10 betas, I love it!! (For those coming from Linucks world, it’s like going from a 2.4 kernel to a snappier 2.6 kernel). Solaris 10 has some amazing breakthrough technologies and amazing capabilities comparing to anything I’ve used till now.
“”This article about something Linux will not get for years (based on the fact that Linux has still not got a half-decent equivalent to kdb, let alone savecore) gets two comments.”
now lets see
http://kgdb.sourceforge.net/
http://www.redhat.com/support/wpapers/redhat/netdump/
Until these are integrated into the kernel, “Linux” still doesn’t have them.
“Until these are integrated into the kernel, “Linux” still doesn’t have them.”
until they are integrated the vanilla source doesnt have them. meanwhile several distributions already have them so for users “linux” does have it.
Until Linus’ tree has them, not all distributions will, so “Linux” does not have them. Red Hat and a few other may, but in general “Linux” does not.
“Until Linus’ tree has them, not all distributions will, so “Linux” does not have them. Red Hat and a few other may, but in general “Linux” does not.
”
if the majority has it then in general linux has it. the vanilla source isnt important for a large majority. in fact distros like redhat have based themselves on the ac branch before. connective is base on the mm branch. so understand what you are talking about
the vanilla source isnt important for a large majority
And which tree do most distributions stay in sync with, Red Hats? Heh.
“And which tree do most distributions stay in sync with, Red Hats? Heh”
the point was end users not developers or distributions. when end users use linux they use it with a crash dump or debugger already enabled and integrated in them so the vanilla source isnt important to them per see. it would better to have them in the vanilla source for various reasons but in general linux for end users already has them so saying that it wont have them or there is no such thing is false
Saying that it isn’t important to end users is really showing just how narrow you POV is. Windows XP for example, not only has such a debugger out of the box, but prompts the user to send information generated from it to Microsoft so as to help fix the issue, improving the software that these people use. Even though they do not have the skills to fix the problem themselves, they still bennefit from it’s inclusion and use.
Now you’re going to say that because I mentioned Windows XP, my arguments are invalid. You’re so predictable.
So DTrace is ‘just’ a way to plug in the D language into certian parts of the kernel (say every time a system call is called?)
I am just trying to understand since they didn’t give a very good definition.
So DTrace is ‘just’ a way to plug in the D language into certian parts of the kernel (say every time a system call is called?)
No dtrace is a dynamic tracing utility, that uses the language D (created specifically for DTrace) to describe what to trace.
Imagine being able to put custom debugging printfs, counters and getting arbitray information such as time taken for a function to execute(and many more such things) of any arbitray function in an appicaltion or in a running kernel, dynamically (no recompiles or reboots) while the system is in priduction. Imagine using a language to describe the information you want and how you want it. D is that language and DTrace is that facility. Imagine being able to do that safely on production systems.
[i] The system had been erroneously configured to be an IPv4 router (i.e., /etc/notrouter and /etc/defaultrouter were both unset), and it would occasionally start to furiously route packets between two other machines related to the benchmark.
You have to be kidding me! so you need Dtrace to weed out poor System Administration?
ou have to be kidding me! so you need Dtrace to weed out poor System Administration?
Yes!!!… mistakes adminstrators make can cause strange things to happen. How would you debug this, hotshot, if you didn’t know the symptoms of this misconfiguration or that it had even occured!!! Support engineers are often called for such problems at customers sites, since they didn’t set the box up and might not have encountered this issue, they have to debug the problem.
Conventional tools would not be able to give you the information you need. Such situations would be precisely the situation for a tool like dtrace.
Raptor said :
“Imagine being able to put custom debugging printfs, counters and getting arbitray information such as time taken for a function to execute(and many more such things) of any arbitray function in an appicaltion or in a running kernel, dynamically (no recompiles or reboots) while the system is in priduction. Imagine using a language to describe the information you want and how you want it. D is that language and DTrace is that facility. Imagine being able to do that safely on production systems.”
Funny how a short comment made me undersrtand what the whole article couldn’t… Thanks a lot Raptor.
The misconfigured E10K system in question belonged to Ebay.
Before they finally discovered the (D’oh!!) misconfigration, Sun had engineers, with sleeping bags, sleeping next the E10K waiting for an event. McNealy was on the phone daily with Meg Whitman giving her status reports.
Talk about service!
Had this been a Windows server I could just hear it now –
Ebay calls MS premium support: Hello our server is all messed up what do we do?
MS: Please hold while I consult our engineers. <Musak in the background>…
MS: Hello, thanks for waiting. Yes, you’re going to need to reinstall Windows.
Ebay: $!@*#.&%$#!@!!
MS: That will be $35.00. Thanks for calling Microsoft premium support.
Yes!!!… mistakes adminstrators make can cause strange things to happen.
I could understand if perhaps someone went crazy with ndd or made some changes to certain system files then perhaps dtrace would be useful..personally if the above situation had occurred i would rebuild the Operation System from scratch. Before you say companies don’t have that luxury of rebuilding systems yada yada remember we are talking about good SA work coupled with a well planned out environment. The particular event the article described was just plain sloppy SA work period and in my opinion a design flaw in Solaris similiar to having filesystem logging off by default. However the defaultrouter file is a well know Solaris quirk.