Anything Forums

This is something I just learned the hard way.

Sure linux in general is more reliable, it won't usually crash for no reason like windows does.

But get this, my server spontaniously ran out of memory, so it just jammed up for no reason, then I went to reboot and it could not do the "scandisk" I had to run this command manually, and now I have tons of screw ups and it ain't done yet.

Simply put, you can say goodbye to Zeroboard, and if something happends to Iceteks, youc an say good bye too, since the backups were ON that drive. Though I have database backups there, and elsewere, but the site itself is there.

Now the perfect setup would be linux on NTFS.

Linux file system is just too unreliable. At least you can shut off windows badly and not loose everything.

Anyway, back to the server, I need to hold down a key so it keeps doing this thing, I might get lucky and not have lost zeroboard but things look really bad right now. I'll probably be switching to a windows based server if this keeps up. It's really too bad since despite being super complicated and user unfriendly, I gave linux a chance, but so many times it just threw it away.

Archived topic from Iceteks, old topic ID:3049, old post ID:25033

That’s really bad!

Were you not using ext3? I was under the impression that it was very reliable and could journalise like NTFS, but I’ve never tried it personally.

Archived topic from Iceteks, old topic ID:3049, old post ID:25034

Nope, I thought the same about EXT3 until this happened... and it ain't the first time, though this time is very bad... Basically as soon as linux shuts down bad, you're screwed. At least in windows it will run scandisk and boot as if nothing happened, even if it was in the middle of something intense like a defrag or file transfer.

Now I see why high end power backup systems are needed for linux servers. My 5 minute UPS just won't cut it during a power outage. Unless I'm there to shut down the server, which is usually the case.

Archived topic from Iceteks, old topic ID:3049, old post ID:25035

I've had plenty of crashes and use EXT3 file systems without any data loss. I suppose it really depends on what you do. Everything should still be there, but there may be some lata loss in individual files. Unless you were performing IO on the files when the server went down, they should be untouched.

I'd look a bit deeper into why the thing went down if I were you. That sounds too odd.

Archived topic from Iceteks, old topic ID:3049, old post ID:25057

I've had windows for about 7 months and i can count on my fingers the amount of times it crashed.

Archived topic from Iceteks, old topic ID:3049, old post ID:25058

Cold Drink wrote: I've had plenty of crashes and use EXT3 file systems without any data loss. I suppose it really depends on what you do. Everything should still be there, but there may be some lata loss in individual files. Unless you were performing IO on the files when the server went down, they should be untouched.

I'd look a bit deeper into why the thing went down if I were you. That sounds too odd.

Yeah well right now I'm in the process of clean installing, since this crash totally fubared everything. All the data is suprisingly still there, but 3 hours of chaos really got me scared, and it's pretty bad for a simple power down at the wrong time. I don't think it was doing I/O at the time either, which is what I find odd. It was basically doing a BSOD but in linux, I went down and it was just a bunch of code and weird out of memory errors but it got jammed there so I hit the 3 key salute, did not do much, so hit reset.

I'm downloading RH9 right now but it's only at 50% and I have another CD to go. So debating wether or not I should just put back 8.0. I want to get this done and over with asap since I need it for my business and other stuff.

Archived topic from Iceteks, old topic ID:3049, old post ID:25061

Hmm using Red Hat? So I guess we found the problem Actualy RedHat is just fine, but you may went to look to CentOS or another RH clone. RH isn't being supported anymore except for enterprise edtion (which I"m sure you're aware of) and that could mean you may not get all the undates you need to get.

I use Gentoo myself, but considering how long it takes to get it up and running, I wouldn't reccomend it if you need your system up, which it seems you do

On a side note, I"m in the process of setting up automated backups to Subversion repositories, the whole lot of which will be compressed and dumped to a Zip 250. You might find it useful

Archived topic from Iceteks, old topic ID:3049, old post ID:25063

Actually I do allot of automated backups in bash which do me just fine, what's nice about cron is that it sends out an email so I have an address set for this.

I got red hat back on the system, sure it's outdated, but think of all the people still using windows 95, that's much worse. Though I'm only starting to configure stuff, the hard stuff has not come yet. Basically the hardest is php/mysql/apache setup, then email, then DNS. Once those 3 beasts are done, I'll be singing in fields of roses. (no, not really )

I was thinking of using a different distro but rather not sacrifice the time to download. I think in the future I'll want to build a low end box, like a celeron or something, so I can test out new distros. I do need to be able to play more with linux without worrying about breaking stuff so a test machine would be great for that.

Archived topic from Iceteks, old topic ID:3049, old post ID:25065

Well, there are two things you can do. A lot of people use Virtual PC or VM Ware and run different Linux distros out of that, so you can try whatever you like.

You *should* be able to do something similar without a comercial product. Depending on the distro's install procedure, you should be able to install to a directory on an existing partition. Then, while still in your "real" distro, you can just chroot in to the new one. Of course you can't overlap services or anything, but you can use XVnc with a login manager to create X in X sessions. Then you can play all you want while keeping your main distro up and stable

Archived topic from Iceteks, old topic ID:3049, old post ID:25067

Yeah that's one way of doing it. What about installing stuff and general file management, would it be contain in it's seperate "space"?

But I think to not complicate things I'll wait till I can afford a seperate box. I priced one for fun and for a low end one it would of costed me about 400 bucks to build which is pretty good. I should try to get a micro ATX case so I can just put it under my bed or something when not in use.

Kind of funny since the athlon 64 chip was cheaper then the 32bit one of lower GHz. My supplier makes me wonder sometimes lol.

Archived topic from Iceteks, old topic ID:3049, old post ID:25068

There is something that will list the last operations that your system was doing before it crashed (linux by default logs everything ), and it's the /var/something/dmesg

I appoligize for the lame help, but there are a few places that have a recording of the problem, and most likely the program that went all buggy should have a log of what happened too.

Archived topic from Iceteks, old topic ID:3049, old post ID:25069

Red Squirrel wrote: Yeah that's one way of doing it. What about installing stuff and general file management, would it be contain in it's seperate "space"?

Everything will be seperate. That's the beauty of chroot Of course, it wouldn't really be a "safe" chroot, but that's not needed. As far as programs, libraries, config, users, etc... go, what's in there stays in there.

If you really want to, you can use loopback devices to mount files as the filesystem for another distro's install. That way, you have much, much lower chance of cross-contamination.

There is something that will list the last operations that your system was doing before it crashed (linux by default logs everything wink.gif ), and it's the /var/something/dmesg

Well it depends. dmesg isn't there all the time, and isn't always that useful for this type of thing. My system has an "everything" log, but in this case I would look at/in /var/log/kern or .../kernel as well as .../daemon and .../messages.

Of course, that wouldn't have helped it if happened to me My syslog daemon buffers writes, so the files aren't up to the second accurate.

Either way its too late now

Archived topic from Iceteks, old topic ID:3049, old post ID:25071

Yeah too late now unless I raw read the drive. But apache was acting up for long I just never got around to troubleshooting it, never realized it was going to actually freeze. My next step is to figure a way to get my ups to shut down the machine properly, in case we ever get a power outage. Would not want this to happen again.

Archived topic from Iceteks, old topic ID:3049, old post ID:25072

http://www.tortuga.com.au/products/upsd/

You should be able to fine a proper RPM out there somewhere for RedHat.

Archived topic from Iceteks, old topic ID:3049, old post ID:25075

Thanks but I already started reinstalling it last night. Got DNS working so far, need to do the rest, hoping to finish by the end of the week.

Archived topic from Iceteks, old topic ID:3049, old post ID:25079

well, when you make manual changes to apache, or any program for that matter, you really need to know what you need and do not need, because if the program is locking up, it might have been something in the httpd.conf file that you misunderstood.

Archived topic from Iceteks, old topic ID:3049, old post ID:25086

No I don't think so anyway. I know httpd.conf pretty well, it's the apache installation string, php, mysql etc stuff that I can confused with but the actual config file for apache I know fairly well, and I had not changed it for a long time before it happened.

Getting samba working right now, hopeing to challange apache/php/mysql either after or tomorow, then email. For email I'll attempt to use maildirs instead of mboxes, that should improve mail processing, since it might even be email that caused the server to crash, but I just happened to refresh a page so apache was the first program to crash from not enough memory. That's a possibility too. Especialy with all the spam rules I have, spamassassin takes a few minutes per email.

Archived topic from Iceteks, old topic ID:3049, old post ID:25087

I used Fedora Core 2 and had problems with unexpected shutdowns and data loss/corruption. I've never had the same problem with Ubuntu Linux, and I've given it a fair amount of stress.

Both cases involved ext3.

Archived topic from Iceteks, old topic ID:3049, old post ID:25091

So I guess it has to do more with the distro then the file system then? Since I did hear Ext3 is just as reliable as ntfs, just have trouble believing after what happened. Just glad I did not loose data, but I did not go through every single file so I might have some corrupted data on my 100GB partition. I'll only know once apache/php/mysql is installed and that I run all my site, if there's parse errors I'll know stuff got corrupted.

Archived topic from Iceteks, old topic ID:3049, old post ID:25092

This is all such a tough subject Filesystem wise, with journaled file systems (NTFS and EXT3) nothing should ever be corrupt, but there may be a mix of old/new data in a file. But that's still only if its being written to. With EXT3, you can specify mount options which acts as a slider between performance and saftey.

Anyway, Im curious as to how much swap space you have, and do you have /tmp mounted at a partition or pointing to memory with tempfs or something. I was going to move my /tmp to tmpfs, but considering how much RAM that would wind up eating...

Archived topic from Iceteks, old topic ID:3049, old post ID:25094

My swap is only 1 Gig, should probably be more... And for /tmp it's just on the same partition as everything else. Should it be on a seperate one?

BTW I got apache and mysql installed, now on to php, and then, making them all work together. It's actually going much smoother then I thought,but I am taking it slowly. Had trouble getting mysql running but now all is good, so it's just the thing of setting it up so it runs at boot time which is not that hard.

Archived topic from Iceteks, old topic ID:3049, old post ID:25097

My swap is only 1G Its a good idea to keep /tmp seperately so no one can eat up your disk space (imagine someone uploading 20 1GB files at the same time through PHP) and because the data there isn't so important, you can mount it as EXT2 or Reiser, and turn off access time updates, etc... for a little extra performance.

Archived topic from Iceteks, old topic ID:3049, old post ID:25099

Oh I see so it's more of a thing that would matter for an online server. So in my case it's not that much of an issue. I guess another way on an online server would be to make a script that periodicly does a DU -hs in that folder and sends an email, so it could be easly monitored.

Archived topic from Iceteks, old topic ID:3049, old post ID:25103

And we got success!

Warning: mysql_connect() [function.mysql-connect]: Access denied for user 'root'@'localhost' (using password: YES) in /data/intranet/main/www/iceteks/scada/mainlogger.php on line 139

Warning: mysql_select_db(): supplied argument is not a valid MySQL-Link resource in /data/intranet/main/www/iceteks/scada/mainlogger.php on line 140
Scada SQL Error: No database selected

Yes, that's success, because it means php mysql and apache are working properly. I just need to restore all the databases from the backups.


Wait a sec... what backups?!?! J/K

Archived topic from Iceteks, old topic ID:3049, old post ID:25106

Red Squirrel wrote: Oh I see so it's more of a thing that would matter for an online server. So in my case it's not that much of an issue. I guess another way on an online server would be to make a script that periodicly does a DU -hs in that folder and sends an email, so it could be easly monitored.

More on how you use the system than specific applications, but yeah.

Right now I'm using about as much resources as I ever do. I've got open a web browser and am compiling in the background, My memory usage is at 256MB RAM (all I have), and about 260MB swap, and 131MB in /tmp. For me, I could get by with /tmp mounted as tmpfs and still have half my swap (1G) left over

This is with:
- Apache 2 allways running
- Another instance of Apache2 allways running for Subversion
- MySQL
- Postfix
- CyrusIMAPd
- Samba
- OpenLDAP (although I never use it...)


Archived topic from Iceteks, old topic ID:3049, old post ID:25107