KnownProblems

From SME Server
Jump to navigation Jump to search

This section is to be used to document problems that cannot or will not be fixed through development of SME Server 7.0. Much of this is obsolete for later versions but left for reference.

Plese be certain any suggestions followed apply to the version you are using!

Backup/Restore

Using a combination of hardware and software compression for tape backup causes errors

You should use only one of software or hardware compression. Doing both may exercise bugs, and won't reduce the size of data on the tape - it is more likely to increase it.


Backup to desktop 4GB limit

As in previous versions of SME Server, the backup to desktop function has a limit of 4GB. It's not fixable until we know where the limit or limits are. All these components are involved in transferring the backup file, and at least one of them is truncating at 4GB: tar, perl, mod_proxy, apache, TCP stack, browser at desktop end. See this bug listing for details bugzilla:701 In SME 7.4 the 4GB limit is effective when using Windows Internet Explorer, however using a browser such as Mozilla's Firefox the 4 GB does not apply and 180 GB has been successfully achieved in a backup-to-desktop. See bugzilla:5222 for more information.

Hardware

This is an incomplete list of problems, you can also check the Hardware_Compatibility_List

Problem with NIC card or integrated NIC

If your NIC card isn't working, replace it with a different one. If your integrated NIC isn't working, disabling onboard lan and inserting pci network card works.

Problem reports have been submitted with the following NIC's:

  • Realtek 8169 Ethernet is not recommended, The driver included in CentOS 5 is out of date and has problems. See Bugzilla:8554 for details on a workaround using a driver package from elrepo.
  • Two identical Realtec nic's caused a kernel panic, Model unknown
  • Intel 82573L Gigabit Ethernet says the e1000 driver should work but doesn't
  • 3com 905 (Cyclone) Ethernet is not recommended. This card 'may' not be detected though a site
has reported success (0000:00:0d.0: 3Com PCI 3c905B Cyclone 100baseTx at 0xb800. Vers LK1.1.19).
  • D-Link DGE-530T Rev-B1 on SME 7.1 does not work.
  • D-Link DFE-530TXS rev. A1 Ethernet adaptor not recognized (a.k.a. DL10050, Sundance) (See also RedHat bug 154349)

Some systems or motherboards may need a bios update or are unsupported

  • v7.5: Latest generation of motherboards do not appear to be compatible (as shipped) with either SME7.5 or 7.6WIP. Bugzilla:6701
  • 2001 CUSI-M motherboard - bios
  • Tyan S2850G2N with AMD Opteron 160, BIOS update >= 112
  • Intel board D946GZIS - unsupported
  • Boards using a Dual Pentium Pro are unlikely to work

SATA II drives are not fully supported

We suggest jumpering sata 300 drives to 150 until there is a newer kernel that supports them and/or the controllers better

Boot options for install

See Booting#Installer Boot Options.

Unsorted

  • Losing too many ticks

A kernel bug with Athlon64 processors may cause the system to run slowly, possibly fixed with a later kernel

  • VGA problem with VIA EPIA board

Once the installer loads the VGA display loses sync and it is impossible to read what is being displayed. There is a workaround, see this bug report for details: Bugzilla:236

  • VIA EPIA CL1000 (dual LAN, single processor)

The smp kernal fails to boot - non-smp works like a charm...

Slow Disk Format

Installing 2 discs as Primary and Secondary on the same IDE channel may cause slow formatting and is not a very good idea because:

  1. May hamper system performance.
  2. Does not provide system redundancy if you loose the IDE channel.

Why does my clock run too fast in SME Server under VMware?

If you are testing SME Server on VMware, you may notice the clock doen't keep time accurately. Perhaps this might be of use: Synaptic: Clock Running fast in CentOS 4.3 [SME7] Guest

Boot failure after install if you have enabled hardware raid

There sometimes seem to be conflicts with the SME Server automatic software raid setup in combination with hardware raid. Disabling the hardware raid in your motherboard or controller card's bios is the suggested fix.

Booting in SMP mode results in kernel panic with certain hardware

The combination of a 2MB Cache P4 3.0GHZ CPU, and a Giga-byte GA-81848P775-G (Intel 848P chipset) motherboard produces this problem. Other simialar hardware may be affected.

Legacy ~MegaRAID driver is no longer included

Certain raid cards that worked under SME Server 6, do not work under SME Server 7.

Adapters with the following PCI vendor ID and device ID pairs are not supported by the megaraid_mbox driver: vendor, device = 0x101E, 0x9010; 0x101E, 0x9060; 0x8086, 0x1960. The lspci -n command can be used to display the IDs for adapters installed in a particular machine. Products with these IDs are known by (but not limited to) the following model names: Broadcom 5820, Dell PERC (dual-channel fast/wide SCSI) RAID controller, Dell PERC2/SC (single-channel Ultra SCSI) RAID controller, Dell PERC2/DC (dual-channel Ultra SCSI) RAID controller, Dell CERC (four-channel ATA/100) RAID controller, DRAC 1, MegaRAID 428, MegaRAID 466, MegaRAID Express 500, HP NetRAID 3Si and 1M.

Both Dell and LSI Logic have indicated that they no longer support these models in the 2.6 kernel. As a result, these adapters are no longer supported in Red Hat Enterprise Linux 4 Update 1, upon which SME Server 7 is based.


Raid on HP server Proliant ML330

Raid on HP server Proliant ML330 with adaptec RAID is unsupported

aic7xxx driver hangs on installation boot

On some Compaq Proliant servers (specifically an ML370 G2), & possibly others, the aic7xxx driver will hang if there is nothing connected to the adaptor or if it is not terminated correctly. On most Proliants you can tell the BIOS to not allocate an IRQ to the device which effectively disables it; the ML370 G2 uses an AIC-7899 chip on the motherboard. This problem also exists in the ~CentOS 4.3 base. SME 6.0 loads fine which would seem to indicate that the newer driver in ~CentOS 4.3 can't handle the older system (this Proliant is probably 5 years old). Bugzilla:373

Anaconda doesn't support >2Tb

Installing on large storage devices will take a significant amount of work. see Bugzilla:1963

Older systems stop at reboot

This may be an issue from the older P2 and early P3 era BIOS'es but the problem may also occur on some newer chipsets, that don't follow the "specs" for power saving features. Also, this problem seems to be more prevalent in multi-CPU machines. Only a kernel change could provide a real fix to this issue. For some motherboards, there may be a workaround by adding appropriate text to the grub config. The addition of the noapic and acpi=off to etc/grub.conf might solve the problem of reboots failing.

Incorrect MD5SUM

I'm burning the SMESERVER cd. When I check the md5sum on this burner, it returns an *incorrect* value. What gives? Try a different cd burner. We've found in the past that some drives provide an extra garbage block when reading back the image. This results in an incorrect md5sum value.

IDE tape drives

I'm trying to use an old tape drive with my SMESERVER but I can't get it working. What gives? SME Server does not provide ide-tape (which is deprecated by Dell and RedHat).

My Adaptec DPT raid controller is not migrated on upgrade. What gives?

The Adaptec DPT raid controller won't work for upgrades. However, the Adaptec DPT raid controller works on new installs. For details visit: bugzilla:833

SME Server installs OK, but freezes on reboot

I've loaded the SME Server on my hardware without problems. Now my SME Server freezes when I reboot. What gives? Most likely this is a problem with the video card or sound card. The best fixes are to replace the video/sound card, disable the video/sound in the bios (if it's integrated on the motherboard) or remove the video/sound card altogether. Video/sound isn't needed on the SME Server and can be managed via the server-manager.

Please note that other hardware devices such as CD-ROM drives as well as mismatched IDE cables may cause a problem, (i.e. kernel panic on reboot after installation). For details, visit: bugzilla:985

kernel: ata1: command 0x35 timeout, stat 0x50 host_stat 0x1

My SME Server suddenly stopped working. I look in the log messages I see: kernel: ata1: command 0x35 timeout, stat 0x50 host_stat 0x1. Any ideas what could be causing this? Most likely, this is either a hardware or BIOS problem or a kernel bug. It's very unlikely we can fix any of those. However, you can try editing /etc/grub/grub.conf. In this file, close to the top, you'll see the latest kernel. It looks like this:

kernel /vmlinuz-2.6.9-42.0.10.EL ro root=/dev/vg_primary/lv_root

Add "noapic" to the end. It should look like this:

kernel /vmlinuz-2.6.9-42.0.10.EL ro root=/dev/vg_primary/lv_root noapic

Then reboot. Hopefully the problem goes away.

Is there a problem installing SME Server with my NFORCE motherboard?

Submitted reports claim that nForce boards based on the 410 or 430 chipset won't work bugzilla:1240.

Is there a problem installing SME Server with my ZIP drive?

Submitted reports claim that installing SME Server on a system with a ZIP drive won't work bugzilla:1348.

Is there a problem installing SME Server with my laptop?

SME Server has never supported laptop PCMCIA network adaptors. This being said, some possible workarounds have been identified in bugzilla:1246

Symbios 53c895 scsi raid card

This is a hardware issue. It worked with SME6 but doesn't work with SME7.

Installation (not hardware related) & Initial Configuration

Why is there a delay after verifying administrator password?

On "slow" system, there is a delay (2-5 seconds) after verifying administrator password. This is because the password is actually set and checked at this point. Everything else we wait until the script is done to perform any actions.

Use /server-manager, not /e-smith-manager

In previous versions, both aliases worked, but in SME Server 7 you must use /server-manager to access the server manager.

Why can the system not be named 'mail', 'ftp' or 'www'?

If you choose one of these for the hostmane of your server, the atalk (Appletalk) service will have errors and fail.

Why is there no option to create reinstallation disk?

There is no option to create a reinstallion disk during installation or later from the Administration section of the server-manager. This is intentional, since it required too much effort to keep it up-to-date as kernels change. The CD contains a rescue mode, which is a better environment for fixing problems, and a reinstall followed by a restore of the configuration works well without us supporting another method. Additionally, the kernel has grown too large for a floppy plus initrd. Even Redhat installs no longer offer to create a boot floppy at install time. More details can be found in bugzilla:792.


4 disk install - md1 is raid1 and md2 is raid5

Even if you have 4 (or more) disks, which are supported by your motherboard and ~CentOS, md1 is always set up as a 4 disk raid1. This is, because you can't boot off a raid-5. md1 is the boot partition and must either be non-raid or raid-1. md2 will be set up as a 4 disk raid5 as expected. More information can be found in bugzilla:759.

Anaconda crash from rescue mode when starting network interfaces

There is a problem with the installer and as networking is not essential for a rescue function it won't be fixed any time soon. Note however that this is no longer an issue with version 7.2. bugzilla:712

On first boot after installation, system name appears on terminal 1 while on terminal 2 and 3 it shows as "localhost".

This is because the gettys on tty2 and tty3 are started before the configuration wizard runs, and therefore before the system name has been set. Also note that the correct system name is displayed after someone logs out from VT2 or VT3?)

The installation gave an exception when trying to format already formatted disks

In some cases, you may have to manually wipe re-used disks.

Services

Restarting NUT

  • My UPS doesn't work although its status says that it's running. What gives?

Rather than:

service nut restart

Use:

service nut stop
<wait a few seconds>
service nut start

Details: bugzilla:2738

Manualy install a service on SME 10 and newer

Starting SME 10 systemd is in used, but there are some backward compatibility for sysvinit and daemontools managed service. Prefered way to install a non core service is by using a contrib, but if you need to run a service without any contrib available, then here are the steps. First check if the rpm providing the service includes either a /etc/rc.d/init.d/servicename or a /usr/lib/systemd/system/servicename.service files. Replace servicename by the appropriate name.

Global steps

Create a db entry in configuration db:

 config set servicename service status enabled

only systemd .service file available

Create a dropin file to have this service available under sme-server.target as this is our target, we do not use the regular multi-user.target.

mkdir -p  /etc/lib/systemd/system/servicename.service.d/
printf "[Install]\nWantedBy=sme-server.target\n" > /etc/lib/systemd/system/servicename.service.d/50koozali.conf
systemctl daemon-reload
/etc/e-smith/events/actions/systemd-default

NB: in a contrib, we would use the /usr/lib/systemd/system/servicename.service.d/50koozali.conf path as it would be provided by an rpm and not by end-user admin.

both systemd .service and /etc/rc.d/init.d/servicename files available

You need to hide the file in init.d from SME Server internals to be able to enable the service form systemd on reboot

mkdir -p /etc/e-smith/templates-custom/etc/systemd/system-preset/49-koozali.preset/
printf "{
\$customservice='servicename';
\$OUT="";
\$OUT .= \"enable                \$customservice.service\n\" if \${\$customservice}{'status'} eq \"enabled\";
}" >> /etc/e-smith/templates-custom/etc/systemd/system-preset/49-koozali.preset/99servicename
expand-template /etc/systemd/system-preset/49-koozali.preset
/etc/e-smith/events/actions/systemd-default

only /etc/rc.d/init.d/servicename file available

cd /etc/rc.d/rc7.d
ln -s ../init.d/e-smith-service S85servicename

only run script in a daemontool runit /service/servicename folder

printf "[Unit]
Description='''servicename'''
Requires=runit.service

[Service]
Type=oneshot
ExecStartPre=/sbin/e-smith/service-status '''servicename'''
ExecStart=/usr/bin/sv u /service/'''servicename'''
ExecStop=/usr/bin/sv stop /service/'''servicename'''
ExecReload=/usr/bin/sv t /service/'''servicename'''
RemainAfterExit=yes

[Install]
WantedBy=sme-server.target" > /etc/systemd/system/'''servicename'''.service
systemctl daemon-reload
/etc/e-smith//events/actions/systemd-default

Packages

Ghostscript ps2pdf and Hylafax

Old contrib pdf over lan stopped working on SME 7, the *.pdf conversion is done with ghostscript ps2pdf. This bug listed on Centos http://bugs.centos.org/view.php?id=5500 also describes issues with Hylafax and ps2pdf.

0005500: new feb 2012 ghostscript-devel-7.07-33.13.el4 RPM for i386 crashes if you use ps2pdf or hylafax

The issue has been introduced by a RedHat security update http://www.redhat.com/archives/enterprise-watch-list/2012-February/msg00008.html. It is unlikel any changes will be coming from upstream to fix this.

A workaround can be initiated by reverting to the last Ghostscript package that supported the feature.

rpm -Uvh --oldpackage ghostscript-7.07-33.11.el4.i386.rpm.

This fixed the issue. Details: bugzilla:6857