• Recent troubles with my Fedora system

    From Lars Poulsen@lars@beagle-ears.com to alt.os.linux.fedora on Sat Aug 1 06:54:12 2026
    From Newsgroup: alt.os.linux.fedora

    I am a 76 year old geek, still working part time in IT. I am part owner
    of a "4 guys in a garage" company doing high-reliability industrial data radios.

    I have always subscribed to the principle of "Windows on the desktop,
    Linux in the back-end", both at work and at home. My Windows desktop
    system at home is a MINIX compact PC (NGC NR660) with an AMD Ryzen5
    6600H with 16GB of RAM and 500 GB of SSD, running Win11 25H2.

    My home Linux box is an an old Dell Precision 3630 with an i7-8700 CPU
    at 3.2GHz with 16GB of DDR4 at 2666 MHz, 1TB SSD and 2x2TB HDD, running
    Fedora Workstation (F43, nightly updates) acting as
    - SAMBA fileserver
    - Apache webserver
    - Postfix post office
    - Dovecot IMAP server
    - DigiKam photo archive (100,000 images)

    My Internet connection is Frontier Business fiber 500GB with a static IP address (that is what makes it business) and an old Netgear edge router,
    which limits me to 100GB - which is more than I need.

    During a recent large home renovation project (mold remediation) the
    Linux box was sitting in my network closet, but kept running. At one
    point, it overheated, and it appears that it got damaged. Every so
    often, it would turn the root file system into read-only. When this
    happened, the email services would fail. It would recover on a reboot,
    but fail again after an interval ranging from 45 minutes to 72 hours. Eventually, it failed to reboot, and I am now rebuilding it.

    I tried hard to find clues to the failure, but when the root filesystem
    - including /var/log - is not writeable, its pretty much a lost cause. I
    was hoping to determine if it was DRAM, CPU or SSD that was going bad,
    but since memtest86+ ran fine, I suspected the CPU. I did however decide
    that a system like this really needs to have ECC memory. Which has now
    shot up dramatically in price.

    As the reboot intervals grew shorter, I had looked for a replacement
    box, and ordered another Dell Precision 3630 with the same CPU. To my surprise, it was not completely identical: It came with an HDMI port in addition to the two DisplayPort jacks. Not that it will do much for me:
    I have a 4K video card in the expansion slot for the benefit of DigiKam.

    When I moved the drives to the new chassis, the system would not boot. I eventually discovered a long forgotten fact: The SSD was formatted with
    MBR instead of GPT. I converted it, but I cannot figure out how to get
    GRUB to pick up my boot menu. Today I will try a new tack: Install the
    system fresh on the SSD from the new box, then add in the additional
    file systems, migrating in the non-root ones from the old SSD last.

    My greatest challenge in doing that, it that the "Live" installer wants
    very much to make the root FS a btrfs. To avoid that, I need to manually
    lay out my partitions, but the partitioning section of the installer
    will not accept the EFI partition I create for it. Some nonsense about
    what to call the partition type: EFI, VFAT or FAT32.

    I feel very lonely struggling with this. I have not found anyone else in
    my town that I can call on to be a second pair of eyes for this kind of
    work. There must be a couple of dozen people with the knowledge, but I
    don't know them socially, and they are probably as old as myself. And
    nobody seems to advertize professional services related to Linux system administration.

    Is that true everywhere?
    --
    Lars Poulsen - an old geek in Santa Barbara, California

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Lars Poulsen@ljp93105@gmail.com to alt.os.linux.fedora,alt.unix.geeks on Sat Aug 1 12:55:29 2026
    From Newsgroup: alt.os.linux.fedora

    [added alt.unix.geeks, because nobody actually uses alt.os.linux.fedora]

    On 8/1/26 06:54, Lars Poulsen wrote:
    I am a 76 year old geek, still working part time in IT. I am part owner
    of a "4 guys in a garage" company doing high-reliability industrial data radios.

    I have always subscribed to the principle of "Windows on the desktop,
    Linux in the back-end", both at work and at home. My Windows desktop
    system at home is a MINIX compact PC (NGC NR660) with an AMD Ryzen5
    6600H with 16GB of RAM and 500 GB of SSD, running Win11 25H2.

    My home Linux box is an an old Dell Precision 3630 with an i7-8700 CPU
    at 3.2GHz with 16GB of DDR4 at 2666 MHz, 1TB SSD and 2x2TB HDD, running Fedora Workstation (F43, nightly updates) acting as
    - SAMBA fileserver
    - Apache webserver
    - Postfix post office
    - Dovecot IMAP server
    - DigiKam photo archive (100,000 images)

    My Internet connection is Frontier Business fiber 500GB with a static IP address (that is what makes it business) and an old Netgear edge router, which limits me to 100GB - which is more than I need.

    During a recent large home renovation project (mold remediation) the
    Linux box was sitting in my network closet, but kept running. At one
    point, it overheated, and it appears that it got damaged. Every so
    often, it would turn the root file system into read-only. When this happened, the email services would fail. It would recover on a reboot,
    but fail again after an interval ranging from 45 minutes to 72 hours. Eventually, it failed to reboot, and I am now rebuilding it.

    I tried hard to find clues to the failure, but when the root filesystem
    - including /var/log - is not writeable, its pretty much a lost cause. I
    was hoping to determine if it was DRAM, CPU or SSD that was going bad,
    but since memtest86+ ran fine, I suspected the CPU. I did however decide that a system like this really needs to have ECC memory. Which has now
    shot up dramatically in price.

    As the reboot intervals grew shorter, I had looked for a replacement
    box, and ordered another Dell Precision 3630 with the same CPU. To my surprise, it was not completely identical: It came with an HDMI port in addition to the two DisplayPort jacks. Not that it will do much for me:
    I have a 4K video card in the expansion slot for the benefit of DigiKam.

    When I moved the drives to the new chassis, the system would not boot. I eventually discovered a long forgotten fact: The SSD was formatted with
    MBR instead of GPT. I converted it, but I cannot figure out how to get
    GRUB to pick up my boot menu. Today I will try a new tack: Install the system fresh on the SSD from the new box, then add in the additional
    file systems, migrating in the non-root ones from the old SSD last.

    My greatest challenge in doing that, it that the "Live" installer wants
    very much to make the root FS a btrfs. To avoid that, I need to manually
    lay out my partitions, but the partitioning section of the installer
    will not accept the EFI partition I create for it. Some nonsense about
    what to call the partition type: EFI, VFAT or FAT32.

    I feel very lonely struggling with this. I have not found anyone else in
    my town that I can call on to be a second pair of eyes for this kind of work. There must be a couple of dozen people with the knowledge, but I
    don't know them socially, and they are probably as old as myself. And
    nobody seems to advertize professional services related to Linux system administration.

    Is that true everywhere?
    --
    Lars Poulsen - an old geek in Santa Barbara, CA
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to alt.os.linux.fedora on Sun Aug 2 09:41:35 2026
    From Newsgroup: alt.os.linux.fedora

    Am 01.08.26 um 15:54 schrieb Lars Poulsen:


    My Internet connection is Frontier Business fiber 500GB with a static IP address (that is what makes it business) and an old Netgear edge router, which limits me to 100GB - which is more than I need.

    The units are wrong.
    It is data per seconds, so 100GB (gigabytes!)/s would be possible, but
    very unlikely for a home connection.
    I assume it is 500 MBit/s and your router only has Fast Ethernet (100
    MBit/), so limits it.

    During a recent large home renovation project (mold remediation) the
    Linux box was sitting in my network closet, but kept running. At one
    point, it overheated, and it appears that it got damaged. Every so
    often, it would turn the root file system into read-only. When this happened, the email services would fail. It would recover on a reboot,
    but fail again after an interval ranging from 45 minutes to 72 hours. Eventually, it failed to reboot, and I am now rebuilding it.

    I tried hard to find clues to the failure, but when the root filesystem
    - including /var/log - is not writeable, its pretty much a lost cause. I
    was hoping to determine if it was DRAM, CPU or SSD that was going bad,
    but since memtest86+ ran fine, I suspected the CPU. I did however decide that a system like this really needs to have ECC memory. Which has now
    shot up dramatically in price.

    Do a badblocks test of all the disks/SSD. If it shows any errors, it is faulty.

    As the reboot intervals grew shorter, I had looked for a replacement
    box, and ordered another Dell Precision 3630 with the same CPU. To my surprise, it was not completely identical: It came with an HDMI port in addition to the two DisplayPort jacks. Not that it will do much for me:
    I have a 4K video card in the expansion slot for the benefit of DigiKam.

    You can monitor CPU temperature and run "stress -c CORES" to test it
    under load.

    When I moved the drives to the new chassis, the system would not boot. I eventually discovered a long forgotten fact: The SSD was formatted with
    MBR instead of GPT. I converted it, but I cannot figure out how to get
    GRUB to pick up my boot menu. Today I will try a new tack: Install the system fresh on the SSD from the new box, then add in the additional
    file systems, migrating in the non-root ones from the old SSD last.

    You have to reinstall GRUB. This depends on whether the system is
    installed in EFI mode or in BIOS mode.

    I recommend using EFI mode. You can even do that on a disk with a msdos partition table (MBR).
    --
    Gru|f
    Marco

    Please send unsolicited mail to dustbin12@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Lars Poulsen@lars@beagle-ears.com to alt.os.linux.fedora on Sun Aug 2 08:32:42 2026
    From Newsgroup: alt.os.linux.fedora

    On 2026-08-02 00:41, Marco Moock wrote:
    Am 01.08.26 um 15:54 schrieb Lars Poulsen:
    My Internet connection is Frontier Business fiber 500GB with a static
    IP address (that is what makes it business) and an old Netgear edge
    router, which limits me to 100GB - which is more than I need.

    The units are wrong.
    It is data per seconds, so 100GB (gigabytes!)/s would be possible, but
    very unlikely for a home connection.
    I assume it is 500 MBit/s and your router only has Fast Ethernet (100 MBit/), so limits it.

    Yes, you are right. But you understood it anyway ;-)

    During a recent large home renovation project (mold remediation) the
    Linux box was sitting in my network closet, but kept running. At one
    point, it overheated, and it appears that it got damaged. Every so
    often, it would turn the root file system into read-only. When this
    happened, the email services would fail. It would recover on a reboot, LP>> but fail again after an interval ranging from 45 minutes to 72 hours.
    Eventually, it failed to reboot, and I am now rebuilding it.

    I tried hard to find clues to the failure, but when the root
    filesystem - including /var/log - is not writeable, its pretty much a
    lost cause. I was hoping to determine if it was DRAM, CPU or SSD that
    was going bad, but since memtest86+ ran fine, I suspected the CPU. I
    did however decide that a system like this really needs to have ECC
    memory. Which has now shot up dramatically in price.

    Do a badblocks test of all the disks/SSD. If it shows any errors, it is faulty.

    I do not think that old-style badblock tests (reading/writing in user
    mode) are valid on modern drives that do automatic behind-the-scenes
    remapping of bad sectors. With these, I am told that only SMART data
    can give useful information. smartctl -a tells me that the SSD is fine.
    The smartctl device database does not have descriptions for a number of
    the SSD specific data fields, but I have added them myself.
    LP>> As the reboot intervals grew shorter, I had looked for a replacement LP>> box, and ordered another Dell Precision 3630 with the same CPU. To my
    surprise, it was not completely identical: It came with an HDMI port
    in addition to the two DisplayPort jacks. Not that it will do much for LP>> me: I have a 4K video card in the expansion slot for the benefit of
    DigiKam.

    You can monitor CPU temperature and run "stress -c CORES" to test it
    under load.

    I never heard of that test before, but it appears to be in a optional
    package called golang-x-tools-stress.

    When I moved the drives to the new chassis, the system would not boot. LP>> I eventually discovered a long forgotten fact: The SSD was formatted
    with MBR instead of GPT. I converted it, but I cannot figure out how
    to get GRUB to pick up my boot menu. Today I will try a new tack:
    Install the system fresh on the SSD from the new box, then add in the
    additional file systems, migrating in the non-root ones from the old
    SSD last.

    You have to reinstall GRUB. This depends on whether the system is
    installed in EFI mode or in BIOS mode.

    Is there is useful how-to for this somewhere? And which GRUB ?
    Apparently there is a GRUB and a GRUB2, and documentation is not very
    clear about whether GRUB2 is a clean replacement for a deprecated GRUB
    or they have different features that are useful in different circumstances.

    (Of course, by the time I got this, I had already blown away my old SSD
    system drive. And by the way, everyone says "do an image backup of the
    whole drive before any significant setup changes", but when the drive is
    a TeraByte, you can spend many hours doing whole-drive copies.)

    After a lot of Googling, I did not find useful information on how to
    hand configure GRUB menus. Most of the how-tos talked about tools that
    worked on the currently running system, not how to do it when you were
    running a "rescue system" (such as a live USB stick) and wanting to set
    up a system drive to be booted later.

    I recommend using EFI mode. You can even do that on a disk with a msdos partition table (MBR).

    The ability of EFI supporting a mix of MBR and GPT drives is not well documented. But in any case, it seems to me that MBR partitioning is now obsolete, because the limitations on partition layout imposed by MBR
    make it hard to use.
    --
    Lars Poulsen - an old geek in Santa Barbara, California
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to alt.os.linux.fedora on Sun Aug 2 20:11:39 2026
    From Newsgroup: alt.os.linux.fedora

    Am 02.08.26 um 17:32 schrieb Lars Poulsen:
    On 2026-08-02 00:41, Marco Moock wrote:

    Do a badblocks test of all the disks/SSD. If it shows any errors, it
    is faulty.

    I do not think that old-style badblock tests (reading/writing in user
    mode) are valid on modern drives that do automatic behind-the-scenes remapping of bad sectors. With these, I am told that only SMART data
    can give useful information. smartctl -a tells me that the SSD is fine.

    Indeed, but it causes the drive to read all sectors currently shown to
    the OS and write to them.
    If that fails, you know the disk is faulty.
    SMART only notices if read errors occur for sectors that are being read. Triggering that is not a bad idea.

    -aLP>> As the reboot intervals grew shorter, I had looked for a replacement LP>> box, and ordered another Dell Precision 3630 with the same CPU. To my LP>> surprise, it was not completely identical: It came with an HDMI port LP>> in addition to the two DisplayPort jacks. Not that it will do much for LP>> me: I have a 4K video card in the expansion slot for the benefit of
    DigiKam.

    You can monitor CPU temperature and run "stress -c CORES" to test it
    under load.

    I never heard of that test before, but it appears to be in a optional
    package called golang-x-tools-stress.

    No, both the command and the package are called "stress".
    It simply creates a process (or multiple) that create 100% load. This
    causes the CPU to run at full clock speed with the maximum heat.
    Then use sensors to check the CPU temperature.

    Same should be done to the GPU using glmark2.

    When I moved the drives to the new chassis, the system would not boot. LP>> I eventually discovered a long forgotten fact: The SSD was formatted LP>> with MBR instead of GPT. I converted it, but I cannot figure out how LP>> to get GRUB to pick up my boot menu. Today I will try a new tack:
    Install the system fresh on the SSD from the new box, then add in the LP>> additional file systems, migrating in the non-root ones from the old LP>> SSD last.

    You have to reinstall GRUB. This depends on whether the system is
    installed in EFI mode or in BIOS mode.

    Is there is useful how-to for this somewhere? And which GRUB ?
    Apparently there is a GRUB and a GRUB2, and documentation is not very
    clear about whether GRUB2 is a clean replacement for a deprecated GRUB
    or they have different features that are useful in different circumstances.

    Yes, but it depends on whether the OS is installed in old BIOS mode or
    EFI mode.
    GRUB nowadays refers to GRUB2 - GRUB 0.x (also called GRUB legacy) is
    out of support and removed from almost all operating system.

    https://docs.fedoraproject.org/en-US/quick-docs/grub2-bootloader/

    (Of course, by the time I got this, I had already blown away my old SSD system drive. And by the way, everyone says "do an image backup of the
    whole drive before any significant setup changes", but when the drive is
    a TeraByte, you can spend many hours doing whole-drive copies.)

    After a lot of Googling, I did not find useful information on how to
    hand configure GRUB menus. Most of the how-tos talked about tools that worked on the currently running system, not how to do it when you were running a "rescue system" (such as a live USB stick) and wanting to set
    up a system drive to be booted later.

    You do not need to do this by hand.
    The GRUB utilities will do that for you.

    I recommend using EFI mode. You can even do that on a disk with a
    msdos partition table (MBR).

    The ability of EFI supporting a mix of MBR and GPT drives is not well documented. But in any case, it seems to me that MBR partitioning is now obsolete, because the limitations on partition layout imposed by MBR
    make it hard to use.

    If you create a new partition table, always use GPT. GRUB can also boot
    from it in BIOS mode.

    Does your current disk have GPT?

    Your computer should already support EFI boot, unless disabled in the
    UEFI setup. Check that.
    Then check if you have a ESP partition (type EF00).
    --
    Gru|f
    Marco

    Junk-Mail bitte an trashcan@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Lars Poulsen@lars@beagle-ears.com to alt.os.linux.fedora on Mon Aug 3 18:38:16 2026
    From Newsgroup: alt.os.linux.fedora

    On 2026-08-01 06:54, Lars Poulsen wrote:
    I feel very lonely struggling with this. I have not found anyone else in
    my town that I can call on to be a second pair of eyes for this kind of work. There must be a couple of dozen people with the knowledge, but I
    don't know them socially, and they are probably as old as myself. And
    nobody seems to advertize professional services related to Linux system administration.

    Is that true everywhere?

    I spent Saturday building a fresh new Fedora system, with the intent of
    then mounting all my non-root drives and getting back in business that
    way. And I did get the system up after a fashion, but rebuilding its
    features is a huge challenge.

    I have samba up and running again, but postfix (SMTP), dovecot (IMAP)
    and httpd (Apache Web server) are still not working.

    Much of this is due to firewall issues. I have for many years used
    hand-edited /etc/sysconfig/iptables, but anaconda keeps insisting that I should be using firewalld, and even though they allegedly use the same backend, the two packages seem to be messing each other up.
    --
    Lars Poulsen - an old geek in Santa Barbara, California
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Marco Moock@mm@dorfdsl.de to alt.os.linux.fedora on Tue Aug 4 08:30:31 2026
    From Newsgroup: alt.os.linux.fedora

    Am 04.08.26 um 03:38 schrieb Lars Poulsen:

    Much of this is due to firewall issues. I have for many years used hand- edited /etc/sysconfig/iptables, but anaconda keeps insisting that I
    should be using firewalld, and even though they allegedly use the same backend, the two packages seem to be messing each other up.

    Indeed, either use firewalld or disable it and manually configure
    nftables. ip(6)tables is still available, but deprecated and nftables is
    used by default as the backend for firewalld.

    firewalld is easy for most situations.
    --
    Gru|f
    Marco

    Junk-Mail bitte an trashcan@stinkedores.dorfdsl.de
    --- Synchronet 3.22a-Linux NewsLink 1.2