Jaguar LOSES speed on my G4 PowerMac

[Numbers1820]

I've searched and searched and not found a thread that seems to address my problem. Which is -

I installed Jaguar on my trusty G4/350MHz/PCI Graphics machine (~800MB RAM) and at first it ran GREAT. Quite a bit faster compared to 10.1.5. I run CPU Monitor all the time in the dock (I love the visual feedback) and for a week or so, the System Processes and User Processes both barely ever nudged over 30% or so. But then I got something I could only label as slowness creep. Doing just "simple" things - opening programs, files, etc. started taking up 50, 70, 99% of the SYSTEM Processes! To give you an example of how slow this is, it took me about 90 seconds to open up the System Preferences pane and select Energy Saver and have it open. Absolutely unexcusable performance!

A bit of history: Under 10.1.5, I noticed a similar slowness creep. In fact, one reason why I hopped on Jaguar was because it was supposed to clean things up. I DID do a clean install - wiped/reformatted my HD, etc. And I still get this....

If this info helps: I have two HDs, the original 10GB HD that I've reserved for OS 9 and a few older files, and a new 80GB that I've partitioned into a 10GB OSX System partition and a 65GB file storage partition. (I wiped the entire 80GB before installing Jaguar, of course.)

I suspect if I knew more about how to interpret the data from Process Viewer or Console or something I could track down this problem, but I don't. (Yet!)

Any ideas, people?

 

[AlanBDahl]

Question: How is performance if you boot in OS9? If you reboot OSX how long does it take for the "slowness" to return?

I'd suggest trying two simple things before looking at more exotic causes. Firstly I'd run Apple's hardware check utility and see if it reports any problems. Secondly I'd suggest running a disk check utility to verify that you don't have a bad sector on one of your HD's.

Have you tried running 'top' from the command-line prompt? I find that it usually gives much more useful information as to what exact process is hogging the CPU.

Good luck!

 

[jocknerd]

type this at the command line:

ps aux

Now scroll up to the top of the listing. The third column will show you the CPU utilization. Scroll down and see if there is a process that seems to be dominating the CPU utiliztion. Something like 50% or higher. If there is something, the second column is the PID (process id). You can kill this process by typing:

kill -15 PID (where PID is the PID listing).

Type ps aux and see if that process was stopped. If not, you may have to type this:
kill -9 PID.

Note: My system's highest CPU usage in a process is currently 9.3%.

 

[drash]

Originally posted by Numbers1820

Any ideas, people?

Man I don't like the sounds of that. Although it sounds like a memory leak, I haven't had a problem with my OS X. More probably it sounds like corruption. Have you tried to run Disk First Aid yet? Or to see if there is a problem on the UNIX side you may have to reboot into Single Use mode by holding down command and s while booting back up (just after you hear the startup bong) and then run fsck -y until you no longer have errors. Type exit when done. HTH

 

[Numbers1820]

Um - what's a memory leak?

 

[drash]

Originally posted by Numbers1820
Um - what's a memory leak?

It's were a program doesn't release memory that it's done with and ends up using all of the available system memory if left running.

 

[shelob]

Type "top -u" in a terminal window, this wil provide you with an active sorted list with the biggest hogs at the top.

Take a look at:

uptime (if the load values are well over 5 then things should be slowing down)

iostat 5 5 (list iostatus every 5 seconds for 5 repetitions)


9.3 % doesn't sound that high, are there a number of jobs running that take around 10%.

 

[Numbers1820]

Originally posted by AlanBDahl
Question: How is performance if you boot in OS9? If you reboot OSX how long does it take for the "slowness" to return?

I'd suggest trying two simple things before looking at more exotic causes. Firstly I'd run Apple's hardware check utility and see if it reports any problems. Secondly I'd suggest running a disk check utility to verify that you don't have a bad sector on one of your HD's.

Have you tried running 'top' from the command-line prompt? I find that it usually gives much more useful information as to what exact process is hogging the CPU.

Good luck!

Ok, this is what I've tried:
I booted into OS 9. Yes, it seemed pretty zippy (although I don't know of/don't have any objective measure like that you find in process viewer or terminal). Then I rebooted into OS X and lo and behold the CPU% overall dropped from an average of 70-90% to something like 30-50% or less. In my experience, for the apps I run (usually more text-based vs. a/v-based, therefore less CPU-intensive), an average usage of 50% or less is just about as fast as I can think, or at least should probably expect out of my 350MHz G4.

I've lost the Apple hardware check disk that came with this G4. The one from my 2001 iBook didn't run on this machine. I did run my iBook's version of AppleCare's TechToolPro and didn't find any issues, although it would not boot from the CD so there were some things it didn't check.

I did boot up from the Jaguar Install CD to run Disk Utitlity/First Aid and everything was verified and repaired. I did not notice any significant performance gains from this.

Last, I did run top from the Terminal and that was pretty revealing. Before booting into OS 9, something called TrueBluEnv (True Blue Environment? what's that, exactly, and is there any way to tweak it?) was hogging anywhere from 40-80% of the CPU cycles. It would cycle up and down every 5 seconds or so. Since booting into 9, then X, it seems to hover around 15% with occasional spikes around 30%. These latter numbers seem to give me acceptable performance.

In sum, it seems that when my system got "gunked up"/slowed down with "high-RPM" CPU cycles (not exactly the precise terms, but this is what it felt like by way of analogy), booting into OS 9 seemed to be the magical remedy. I'm amazed at how so many things in computerland seem to work magically when you stop/restart/shut them down then restart them. There's some basic principle there that i can't quite grasp but it sure does seem to work. The trick seems to be able to identify the right thing to stop/restart....

Thanks for your help!

 

[Numbers1820]

Originally posted by jocknerd
type this at the command line:

ps aux

Now scroll up to the top of the listing. The third column will show you the CPU utilization. Scroll down and see if there is a process that seems to be dominating the CPU utiliztion. Something like 50% or higher. If there is something, the second column is the PID (process id). You can kill this process by typing:

kill -15 PID (where PID is the PID listing).

Type ps aux and see if that process was stopped. If not, you may have to type this:
kill -9 PID.

Note: My system's highest CPU usage in a process is currently 9.3%.

Ok, not yet being quite the Terminal wiz I'd like to be someday, I was real scared to shut down things like TrueBluEnv, let alone top and similar essential-looking processes. What would have happened if I had? Any way to tell what is essential and what is not?

 

[Numbers1820]

Originally posted by jocknerd


Note: My system's highest CPU usage in a process is currently 9.3%.

Oh, BTW, as you can tell from my reply to others, my current highest-usage processes are running something around 10-20%. I suspect my relatively higher numbers are due to what is probably my relatively lower-performing CPU/RAM, though you don't seem to list the specs for your system...

 

[Captain Code]

TrueBluEnvironment is the Classic environment.

You can kill it if you aren't needing Classic, or shut Classic down in the System Preferences.

 

[Numbers1820]

Originally posted by drash


Man I don't like the sounds of that. Although it sounds like a memory leak, I haven't had a problem with my OS X. More probably it sounds like corruption. Have you tried to run Disk First Aid yet? Or to see if there is a problem on the UNIX side you may have to reboot into Single Use mode by holding down command and s while booting back up (just after you hear the startup bong) and then run fsck -y until you no longer have errors. Type exit when done. HTH

As you can tell from another reply, Disk First Aid didn't seem to turn up anything significant.

I did try that Single Use[r?] mode and boy, was that freaky! Seeing my Mac come up with a real terminal (with a small "t") black screen w/ white characters was like something I hadn't had to deal with in 15 years or so! But I did type in fsck -y and it didn't turn up any errors whatsoever. In fact (this may not be the exact quote), the response I got was, "OS X System seems to be ok." ("OS X System" is what I've named the 10GB partition on my add-on 80GB HD, and is where I've been installing the various flavors of OS X as they've come out.) So, this might be a helpful hint for another problem in the future, but it didn't seem to dig up any dirt this time, which is good news, I suppose.

 

[Numbers1820]

Originally posted by drash


It's were a program doesn't release memory that it's done with and ends up using all of the available system memory if left running.

Ok, it seems like this is something I could spot easily using top in Terminal or Process Viewer, right?

The question that I ran into (if you've noticed my replies, above), is that I did NOT find any 3rd-party apps misbehaving in this way, only essential-looking, root-ish stuff like TrueBluEnv. If you have anything to add, like how to tell that killing what process will really screw me up and which ones are just being rude and need to be put down, well, hey, I'm listening.

 

[Numbers1820]

Originally posted by shelob
Type "top -u" in a terminal window, this wil provide you with an active sorted list with the biggest hogs at the top.

Take a look at:

uptime (if the load values are well over 5 then things should be slowing down)

iostat 5 5 (list iostatus every 5 seconds for 5 repetitions)


9.3 % doesn't sound that high, are there a number of jobs running that take around 10%.

I got to your suggestions last (like in the last 1/2 hour or so), but this is what I found for uptime:
[g4350:~] brad% uptime
11:30AM up 1:51, 4 users, load averages: 1.09, 1.12, 1.03

From your advice, it looks like load averages [= values?] are well, below 5, which I guess is a good thing, right?

And here's what iostat 5 5 revealed:
[g4350:~] brad% iostat 5 5
iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics
disk0 disk1 cpu
KB/t tps MB/s KB/t tps MB/s us sy id
11.88 3 0.03 18.51 11 0.20 32 26 42
0.00 0 0.00 0.00 0 0.00 20 15 65
0.00 0 0.00 4.75 1 0.00 12 14 74
0.00 0 0.00 0.00 0 0.00 6 10 84
0.00 0 0.00 0.00 0 0.00 7 11 82
[g4350:~] brad% iostat 5 5
iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics
disk0 disk1 cpu
KB/t tps MB/s KB/t tps MB/s us sy id
11.87 2 0.02 18.50 8 0.14 32 23 45
0.00 0 0.00 0.00 0 0.00 27 16 57
0.00 0 0.00 0.00 0 0.00 24 16 60
0.00 0 0.00 0.50 0 0.00 22 15 63
0.00 0 0.00 0.00 0 0.00 21 14 65

These iostat readings were taken about 20 minutes apart (the time that I've been busy replying to all these helpful hints!).

I'm guessing that the disk 0 and disk1 tables are measuring my HD performance, which I don't think is the issue, right?

Then there are three columns under the cpu heading, us [user?], sy [system?] and id [unused?]. I guess that the higher the id number, the more efficient my system is, and the higher the us/sy numbers are, the more bogged down it is, right? Anything else I should notice, interpret, and act upon from this data?

Now a message for all of you guys: Thanks for the terminal and diagnostic tips! I usually find that a solution is simple in hindsight, but asking people like you sure saves me a lot of time and hair-pulling.

 

[hypocampers]

For passive usage, i.e. not running anything else but what you've booted into, values over 1.00 are very unsual, you should see orders of magnitude less than that, i.e around 0.1 - 0.2 on a G4 933Mhz

When you looked at the output from "top -u" what were the names of the biggest hogs?

I am not too happy about:
"iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics"

I wonder if there are some libraries missing in the installation.

Maybe you can trap a page of "top -U" output and include that in the next post, that is when you are experiencing "slowness".

The "top -u" ouput in Apples rendition seems to display more than the Solaris stuff I've been used too.

From the iostat it looks like later on your machine is spending a lot more time on user processes.

What machine are you running, memory, the status of your root disk (df -k), maybe the machine's disk is getting full and is swapping to too little disk space.

We shall be doing what SUN do shortly and asking for "explorer" logs. Actually that may not be a bad thing to put together for the Mac platform, they are a collection shell scripts, that enable remote support to have a handle on a goodly bunch of aspects of your machine.

 

[drash]

Originally posted by Numbers1820

But I did type in fsck -y and it didn't turn up any errors whatsoever. In fact (this may not be the exact quote), the response I got was, "OS X System seems to be ok."

Excellent. But from your previous post with "TrueBlueEnv" I believe is what is pegging it (I should have asked whether Classic was running). Try making an alternate boot-up extension set for Classic that doesn't enable anything - like take out the USB and network stuff. You don't need it because OS X will take care of that. You might also want to take out extensions that are not normally part of the startup like Adobe Type Manager and anything to do with Norton . Mind you, I'm just talking about OS 9 stuff.

 

[Numbers1820]

Originally posted by hypocampers
For passive usage, i.e. not running anything else but what you've booted into, values over 1.00 are very unsual, you should see orders of magnitude less than that, i.e around 0.1 - 0.2 on a G4 933Mhz

When you looked at the output from "top -u" what were the names of the biggest hogs?

I am not too happy about:
"iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics"

I wonder if there are some libraries missing in the installation.

Maybe you can trap a page of "top -U" output and include that in the next post, that is when you are experiencing "slowness".

From the iostat it looks like later on your machine is spending a lot more time on user processes.

What machine are you running, memory, the status of your root disk (df -k), maybe the machine's disk is getting full and is swapping to too little disk space.

Let me try to answer your questions in nearly reverse order. First (um, I thought I included this info before, but...), these problems are on a first-gen G4, i.e., "PCI Graphics" G4350MHz. I have 800+MB RAM; should be plenty for what I do. I'm not surprised your G4933 runs circles around my 350.

Here is my df -k reading:
Last login: Wed Sep 18 10:57:29 on ttyp2
Welcome to Darwin!
[g4350:~] brad% df -k
Filesystem 1K-blocks Used Avail Capacity Mounted on
/dev/disk1s9 10239040 5950572 4186080 58% /
devfs 96 96 0 100% /dev
fdesc 1 1 0 100% /dev
<volfs> 512 512 0 100% /.vol
/dev/disk0s5 10020816 6136652 3884164 61% /Volumes/Original HD 10 GB
/dev/disk1s10 67903680 9258872 58644808 13% /Volumes/65GB Workspace
automount -fstab [343] 0 0 0 100% /Network/Servers
automount -static [343] 0 0 0 100% /automount
[g4350:~] brad%

And here are my last two iostat 5 5 readings:
[g4350:~] brad% iostat 5 5
iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics
disk0 disk1 cpu
KB/t tps MB/s KB/t tps MB/s us sy id
11.84 0 0.00 18.64 1 0.02 9 14 77
0.00 0 0.00 0.00 0 0.00 31 38 31
0.00 0 0.00 0.00 0 0.00 31 36 32
0.00 0 0.00 0.00 0 0.00 31 39 30
0.00 0 0.00 0.00 0 0.00 30 39 30
[g4350:~] brad% iostat 5 5
iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics
disk0 disk1 cpu
KB/t tps MB/s KB/t tps MB/s us sy id
11.79 0 0.00 18.68 1 0.02 9 14 77
0.00 0 0.00 19.80 20 0.38 71 27 2
0.00 0 0.00 17.63 37 0.64 64 32 4
0.00 0 0.00 24.58 20 0.49 69 29 3
0.00 0 0.00 15.19 26 0.38 67 32 1
[g4350:~] brad%

And top -u:
Um, well, that's a dynamic stat, but glancing over my browser window here, it looks like Software Update (I'm installing 10.2.1 as we speak!) is hogging up about 60% of the CPU, followed by top with around 10%, and everything else (including MS IE) 3% or less.

Oh, Software Update just notified me that it wants me to restart... Here's my latest iostat 5 5 reading:[g4350:~] brad% iostat 5 5
iostat: sysctl(kern.tty_nin) failed: No such file or directory
iostat: disabling TTY statistics
disk0 disk1 cpu
KB/t tps MB/s KB/t tps MB/s us sy id
11.76 0 0.00 18.88 1 0.03 10 14 77
0.00 0 0.00 0.00 0 0.00 32 24 43
0.00 0 0.00 0.00 0 0.00 24 25 50
0.00 0 0.00 0.00 0 0.00 23 28 49
0.00 0 0.00 0.00 0 0.00 24 32 44
[g4350:~] brad%

Well, even when the id value was really low (like single digits), my system felt fairly responsive. I think that must be because the system never seemed to take more than 39%.

Ok, I'm going ahead and restart my machine and see if 10.2.1 makes any difference... Thanks for your thoughtful reply in the meantime!

 

[hypocampers]

This was to check that the root partition and in this case disk "/" was not full, if full then the swap would be more or less useless, but the swap process would continue to thrash to the detrement of the system.

Filesystem 1K-blocks Used Avail Capacity Mounted on
/dev/disk1s9 10239040 5950572 4186080 58% /

Keep us upto date with your update and performance. From iostat the USER part looks to high, so the system doesn not look at fault.

Some network setups (DHCP or static IP etc) can cause system problems, doesn the system seem just to slow or does it have a few seconds where nothing happens at all. The latter could well mean a network setup that he machine is looking for but there is no network support for.

Your machine looks fine, I'm running quite happily on a G3 500, although its does have 512Mb RAM, RAM is always good.

Where did you get the 10.2.1 update, oops I just turned my machine on and its there.

 

[Numbers1820]

Originally posted by hypocampers
This was to check that the root partition and in this case disk "/" was not full, if full then the swap would be more or less useless, but the swap process would continue to thrash to the detrement of the system.

Filesystem 1K-blocks Used Avail Capacity Mounted on
/dev/disk1s9 10239040 5950572 4186080 58% /

Keep us upto date with your update and performance. From iostat the USER part looks to high, so the system doesn not look at fault.

Some network setups (DHCP or static IP etc) can cause system problems, doesn the system seem just to slow or does it have a few seconds where nothing happens at all. The latter could well mean a network setup that he machine is looking for but there is no network support for.

Your machine looks fine, I'm running quite happily on a G3 500, although its does have 512Mb RAM, RAM is always good.

Where did you get the 10.2.1 update, oops I just turned my machine on and its there.

1. I assume that "58%" is not too full for an HD? ...that everything else on this particular test looks ok? (I really don't know how to interpret the data.)

2. Are you saying that "User" includes "network setups"? FYI, I have DSL, linked to a router connecting my desktop and laptop, DHCP. Hmm...it seems that a lot of the slowdowns started happening (not all at once, but incrementally) AFTER I installed DSL w/in the last few weeks! Actually, the order of installation was this: 10.1.5 (round about late spring or whenever it came out; system seemed to run fine); DSL (w/in the last month; system seemed to run fine, THEN "slowed down" to an absolutely unacceptable crawl; I didn't bother to try to troubleshoot this because I knew that Jag was coming out and I figured that'd just fix things ); 10.2 in late August (system seemed EXTREMELY snappy (a miracle!) for about 1-2 weeks, then the slowdown that moved me to post a thread here).

3. Hmm #2...would any of Apple's included OS X network utility apps tell me anything useful? Terminal processes?

4. I'm not sure if I know what you mean by "seem just to slow" vs. "nothing happens at all". Probably the latter has been my experience. But that could be because the response was VERY slow.

5. If something in my network is the problem, why has it affected the G4 only, and not my G3 iBook? My 'Book has nearly always been a really zippy little buddy for me.

6. Thanks for your offer to keep you up to date with iostat reports! I'll monitor my G4's performance, both subjectively and by keeping an eye on CPU Monitor, and if it seems to change much one way or another, I'll post a new iostat or whatever else seems pertinent here.

 

[hypocampers]

1. Your root disk is ok at 58%, but more importantly you have 4Gb available. Your disk is 10Gb, youve used around 6GB and 4 GB remain.


2. Network events take the highest priority within a system as network traffic has to be deal with "right now" or it gets missed, disk stuff can be made to wait etc. So what happens is that a packet going out is being waited upon for an aknowledgement, however if for some reason nothiing comes back for ages the the machine can appear to hang "stop and start". This usually happens as a result of the IP settings on your computer (IP manually set, DHCP, DHCP with manualy setting, etc) , the routing or the Ethernet setting on the Mac and/or router.

I don't have a switched port at home so I can't test the autonegotiation, but many problems in the pat have been caused as a result of Full/Half Duplex problems running at either 10/100 Mbs. The G4 may have a Gigabit port.

Anyway in that instance I set the port on the switch DSL modem (I am taking it as read that you use the DSL modem as the hub) to an exact setting, i.,e. 100fdx and set the Ethernet port on the Mac to match that. Dont' as ke me just now houw to set that speed and duplex on the Mac at this point as I don't know yet.

On Sun, SGI and IBm kist its failry easy, either use a script, or a one time manual setting.

3. The description suggests something a little odd is going on, on one hand as the problems seem to get worse over time.