Public timeline

Notices

1. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Sunday, 31-May-2026 21:41:36 EDT Clarissa Walker
   Updates!
   
   mariadb: 11.4.10 -> 11.4.12
2. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 29-May-2026 14:55:07 EDT Clarissa Walker
   Slowly collecting my custom sensors3.conf files for the following motherboards:
   
   Intel SAI2
   Tyan Tomcat S2925/S2925-E
   ASUS P8H61-I R2.0
   HP Pro 3500
   Supermicro X9SCM-F
   
   
   The ones that are pretty much finished are the Tyan Tomcat S2925/S2925-E and the Supermicro X9SCM-F.
   
   The Intel SAI2 one is very messy and should be completely redone, the HP Pro 3500 one is preliminary and should not be used, and the ASUS P8H61-I R2.0 one should be cleaned up a bit, but is usable otherwise.
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 30-May-2026 14:40:54 EDT Clarissa Walker

     in reply to
     Forgot the MSI PRO Z690-A WIFI DDR4. (DDR5 and any non-WIFI versions as well.) However, it uses a third-party kernel sensors module, so I may not submit that one... and the original in-kernel module is... not all that great.
     
     Also, I should set up the ASRock J4125B-ITX for its sensors.
3. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Thursday, 21-May-2026 11:44:38 EDT Clarissa Walker
   Finally found out why the Apache httpd service would fail at boot: this server is dual-stack, and the web server's service would try to start before the IPv6 part of the network connection would be established.
   
   --
   
   httpd.service:
   * Added network-online.target to After=, didn't work;
   * Added ExecStartPre=/bin/sleep 5 worked, but that just makes httpd delay start for 5 seconds...
   * ...so Restart=on-failure, RestartSec=1s, and StartLimitInterval=1min were to be added instead as a workaround.
4. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 13:18:50 EDT Clarissa Walker
   More changes to the scripts:
   
   * httpd.conf now has the mod_systemd module added and enabled (missed in the initial tidying up process for httpd 2.4.66).
   * httpd.service now properly supports systemd handling. ExecStart is also changed: FOREGROUND option now used.
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 13:26:23 EDT Clarissa Walker

     in reply to
     Goofed on the httpd.service file: KillSignal should not be mixed!
     
     KillSignal changed back to SIGCONF, KillMode should be mixed...
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 13:27:25 EDT Clarissa Walker

     in reply to
     Rather... SIGCONT...
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 13:30:17 EDT Clarissa Walker

     in reply to
     httpd.service: Unit : After now has network.target added. Hopefully this fixes the 'fail-at-start' after a system reboot.
5. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 11:05:11 EDT Clarissa Walker
   Updates!
   
   Security
   httpd: 2.4.66 -> 2.4.67
6. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 15-May-2026 10:30:58 EDT Clarissa Walker
   Welp, Apache httpd needs to be updated due to security issues...
7. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Tuesday, 05-May-2026 00:46:40 EDT Clarissa Walker
   More tidying up the (quite) ancient scripts:
   
   * httpd.service no longer uses the apachectl script and now uses the httpd executable instead.
   * the netplan 00-installer-config.yaml file had a forgotten change for one of the network adapters.
   
   For the network setup, it's quite jank: IPv4 is handled by netplan and IPv6 is handled by systemd-network. This is due to a historical (2017) setup. Initial IPv4 migration to systemd-network actually broke this a few years ago (around 2019), and I never attempted to fix this since. It should be possible now, though. I also learned to not uninstall netplan at that time...
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Tuesday, 05-May-2026 01:15:21 EDT Clarissa Walker

     in reply to
     A bit of Ubuntu Server network config history:
     
     Ubuntu Server, like a lot of distros, stored network configurations under a single filename: /etc/network/interfaces. Of course, starting and stopping them was a bit different.
     
     With version 17.04, they changed it to netplan. Both IPv4 and IPv6 are handled there, and configs were stored under the /etc/netplan directory. Starting and stopping network interfaces changed as well.
     
     With version 19.04, they changed it again to systemd-network. They helpfully migrated the configs to the appropriate /etc/systemd/network directory and had 'netplan' in the filenames. (Uninstalling netplan would be a bad idea here; you need renderer: networkd in a netplan .yaml file instead.)
     
     --
     
     That same year, I spent over 30 minutes attempting to migrate IPv4 over (IPv6 was successfully migrated and was disabled in netplan) but eventually gave up, mainly because I knew no better of the systemd-network formatting back then. So this pretty jank network interface setup that stayed for six years is a result of the double migration of interfaces.
8. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 02-May-2026 15:35:10 EDT Clarissa Walker
   The system upgrade is finished! I may have disabled the onboard video to make it headless on the OPNSense box, as now booting up the webserver will bring up the beep codes for that... maybe next time.
   
   Also, found out that the old swap partition was corrupted, so that was fixed.
   
   It's now possible to run future versions of MariaDB. However, version 11.4 will still be used for now.
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 02-May-2026 15:55:56 EDT Clarissa Walker

     in reply to
     Forgot to mention: the HDD -> SSD swap is probably the main factor of cutting down power, the CPU/motherboard swap being second.
     
     I shaved off ~40W power usage by doing both.
9. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Sunday, 26-Apr-2026 16:47:51 EDT Clarissa Walker
   Okay, ordered an SK Hynix SC311 128GB 2.5" SATA SSD for the replacement board.
   
   This actually delays the migration by about a week. It may also need a firmware update - specifically, a may not occasionally boot scenario.
   
   I also had an idea: relocating the database to another drive... again. I did this last time and it was a success, but it was with 2x 80GB IDE drives as a dm-integrity RAID 1 test.
   
   The database is on the same drive setup as the rest of the webserver data... which is not the system drive.
   • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Friday, 01-May-2026 15:32:06 EDT Clarissa Walker

     in reply to
     The drive arrived today, and it's pretty abused. ~53% life left. This is fine to me, since this drive will not get a lot of writes from me, anyway.
     
     My current Silicon Power MLC SSD's life left is 32%. 😏
     
     Anyway, it also sorely needed a firmware update. This fixes the following:
     - issue where the solid-state drive (SSD) does not resume from low-power mode
     - issue where the system fails to boot when it is not turned on for a long period
     
     I also updated the firmware of a Toshiba MQ01ACF050 500GB 7200RPM drive due to the following:
     - Improved the shock handling capability of the hard drive in power-on state.
     
     This server will be offline May 2, 2026 for the system migration.
10. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 27-Apr-2026 18:24:42 EDT Clarissa Walker
    Follow up to the notice below, I was right, it was the Sony display.
    
    I touched up the diodes in the battery circuit on the Supermicro board, thoroughly cleaned it up, and powered it on. The display happened to glitch up at that time, so instead of rebooting the system, I turned the display off and back on again... and the display was normal again.
    
    I have two of these 20+ year old displays, and this panel is the first of the two... and it happens to have rare display issues, but the panel has no dying pixels, and the bezel is fine. It also has a matching back. The second one has (barely noticeable) dying pixels, but it functions perfectly otherwise... but it has a dent in the top right corner. It also has no matching back.
    
    I'm going to swap the power board and Tcon from the second one into the first one and retire the other display. Dad has the second one (which replaced the first), so it goes back to him after this.
    
    I'll have to retrieve another display for the retro setup, it seems.
    
    http://cwcyrix.nsupdate.info/gnu-social/public/notice/232
11. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 25-Apr-2026 11:33:54 EDT Clarissa Walker
    The replacement server motherboard has arrived. I haven't tested it just yet, however.
    
    Going to have to obtain a SATA drive, though. The board can also use a SATA DOM (Disk-On-Module), and those are usually of industrial spec.
    • Show all 7 replies
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 25-Apr-2026 16:01:47 EDT Clarissa Walker

      in reply to
      After the CPU repaste: http://cwcyrix.nsupdate.info/gnu-social/public/url/54
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 25-Apr-2026 16:03:22 EDT Clarissa Walker

      in reply to
      Yes, I have actually brought CPU temps down nearly 20° Celsius because the original thermal paste was bone dry.
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Saturday, 25-Apr-2026 18:48:20 EDT Clarissa Walker

      in reply to
      And... it also seems that I finally managed to get the board to keep its BIOS settings.
      
      That literally took over an hour to solve that.
      
      It originally started with it not responding with the power button, then eventually I walked off to retrieve something (over two minutes), then tried again and realized it powered on...
      
      Set the time and date, exit, power off the board then the PSU... wait ~2 minutes, power on the PSU, then the board. Forgot its settings.
      
      Repeat this ad-nauseam between testing RAM for errors and CPU temp. Oh, and the display would glitch up at times. May have to retire that particular old Sony SDM-HS94P monitor.
      
      Earlier, I disabled the BMC by jumper; the BMC was still zombie-enabled, but now the BIOS would no longer acknowledge it. Odd.
      
      Eventually jumpered the Chassis Intrusion setting and then used the Set User Default settings in BIOS. From there, BMC was no longer running...
      
      The board is in standby; it needs a SATA drive before the system migration.
12. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Wednesday, 22-Apr-2026 19:56:14 EDT Clarissa Walker
    Well, this future upgrade had me going through my archived stuff regarding the Supermicro X9SCM-F mainboard, since the OPNSense firewall uses one. I should turn off the IPMI/BMC on that, since I have not used it in over five years... It's not even hooked up, either. It merely adds ~2 minutes of boot time to the entire process...
    
    Anyway, found some missing BMC firmware and one newer BMC firmware on the Internet. I am also obtaining the rest of the driver CDs, since an earlier attempt was done when I didn't even have enough space in general.
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Wednesday, 22-Apr-2026 20:14:35 EDT Clarissa Walker

      in reply to
      Forgot to mention: found one questionable file (possibly slightly corrupt). It was a driver CD image. Moved that one to 'questionable'.
      
      One BCM/IPMI firmware was repacked and a few third-party source files were loose, so I repacked them into RAR files.
      
      There actually was one official BMC/IPMI .zip file that was packed a little over a week before the known good checksum one, so it was moved to 'old'. The files themselves matched the ones in the known good one, too.
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Wednesday, 22-Apr-2026 21:22:55 EDT Clarissa Walker

      in reply to
      Found out the other old drive CD image is questionable as well. Going to have to re-download it. So both 2021-era driver CD image downloads have mismatched checksums...
13. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 20-Apr-2026 12:47:38 EDT Clarissa Walker
    Found and ordered a motherboard + CPU + RAM combo to replace the current build this very webserver is using.
    
    Motherboard: Supermicro X9SCM-F
    CPU: Intel Xeon E3 -1230 v2 (3.30 GHz)
    RAM: 16GB DDR3 (when it arrives, I will note the exact specs)
    
    
    Current setup is the following:
    
    Motherboard: Tyan Tomcat S2925 (with S2925-E v2.01 BIOS)
    CPU: AMD Phenom II X4 945 Rev. C3 (3.00 GHz)
    RAM: 8GB DDR2 (HP marked, uses Micron chips)
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 20-Apr-2026 12:48:51 EDT Clarissa Walker

      in reply to
      For fun, here is a comparison (the 125W TDP is actually wrong on the Phenom II; it's 95W):
      
      https://www.cpubenchmark.net/compare/1189vs7/Intel-Xeon-E3-1230-V2-vs-AMD-Phenom-II-X4-945
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 20-Apr-2026 13:11:22 EDT Clarissa Walker

      in reply to
      I was just made aware that the Rev. C2 version of the AMD Phenom II X4 945 can be 125W TDP (most are 95W TDP), so that part is correct. So it does not account for the 95W Rev. C2 or C3 version of the CPU...
14. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Sunday, 19-Apr-2026 21:41:21 EDT Clarissa Walker
    Finally found the reason the MariaDB 11.8 releases do not run on this quite old AMD server: the AM2/AM2+ architecture is too old. (It's missing the PCLMULQDQ instruction flag.)
    
    I went through the trouble of setting up a test server on a MiniPC (Intel NUC NUC6AYH), and it does run there. Though, I forgot to run lscpu for the instruction flags... so I searched for it online, and that NUC's CPU does indeed have the flag.
    
    I then actually compiled the source on an even older Dell Latitude D830 and executed it. Same illegal instruction (SIGILL) as on the aging server. It does not have the flag.
    
    It does run on my main laptop that has a high-end Sandy Bridge CPU, which has the flag.
    
    --
    
    From the AMD server and main laptop logs:
    
    MariaDB 11.4: InnoDB: Using generic crc32 instructions
    MariaDB 11.8: InnoDB: Using crc32 + pclmulqdq instructions
15. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Thursday, 16-Apr-2026 02:47:24 EDT Clarissa Walker
    Okay, MariaDB has updated config from 'medium' to 'large' settings (cribbed from MySQL 5.5 sample configs).
    
    I archived those configs for reference.
16. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Thursday, 16-Apr-2026 01:01:13 EDT Clarissa Walker
    Was going through the server and deleting very old versions of compiled daemons and libraries; freed up ~5GB of space on the boot drive so far.
    
    However, I found an old .tar image of the server 'www' data from September 6, 2023, taking up ~548 GiB on the www-data RAID setup. (The final one is .xz'ed, taking up ~449 GiB on the NAS.) Thought I purged it years ago, lol.
    
    There's also some old HP laptop disk image on there from 2014 that I have never removed, taking up ~109 GiB; since copied to the NAS years later.
    
    Another ~800 GiB used up there is a bunch of Teknoparrot ROMs just sitting there as well.
    
    Very old versions of MySQL sit there; no longer can use them since the move to MariaDB.
    
    There's more decade [plus]-old cruft on that setup that existed ever since the server used a 4x 250GB RAID0+1 setup in the early 2010s. Removing a lot of it would free up probably ~2TiB of space.
17. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Wednesday, 15-Apr-2026 21:43:22 EDT Clarissa Walker
    Updated the structure of the *very* old httpd.conf file, as it was using the layout from version 2.4.33. This also makes my custom mime.types file finally obsolete.
    
    Also rectified the MariaDB update oversight, as I actually thought I did so yesterday...
18. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Tuesday, 14-Apr-2026 13:56:09 EDT Clarissa Walker
    Updates!
    
    openssl: 1.1.1w -> 1.1.1zg
    mariadb: 11.4.7 -> 11.4.10
    
    Recompile:
    
    httpd 2.4.66 - due to external pcre2 and openssl updates
    php 7.4.33 - due to external pcre2 and openssl updates
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Tuesday, 14-Apr-2026 13:59:40 EDT Clarissa Walker

      in reply to
      Somehow, one of the PHP -FPM configs wouldn't stick (I changed the listen user/group entry to what httpd is using and it was stuck on default nobody...). I had to edit it locally and not from my laptop. 🤨
19. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Sunday, 12-Apr-2026 16:40:29 EDT Clarissa Walker
    Updates!
    
    httpd: 2.4.62 -> 2.4.66
    
    Consistency Fixes...
    
    php 7.4.33: fpm now consistently runs on Unix sockets instead of TCP/IP:Port, meaning connections within PHP are even faster. This was mainly due to merging out-of-date configurations that I thought were up-to-date. [Was originally CGI for the longest -> FastCGI for testing -> FPM TCP/IP:Port -> FPM Unix Sockets]
    
    Other notes: MySQL/MariaDB did run on TCP/IP:Port due to Unix Sockets breaking at random historically, but that had been stabilized for at least a year.
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 13-Apr-2026 23:15:20 EDT Clarissa Walker

      in reply to
      Forgot:
      
      apr: 1.7.4 -> 1.7.6
20. Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 23-Mar-2026 13:01:32 EDT Clarissa Walker
    It seems that since around Linux kernel 6.6, newer Intel wireless cards will crash at suspend/wake... that is, unless you turned off the radio of the card entirely before suspending the system. My main laptop is still using Linux 6.2 (from 2023) due to this. I previously had 6.5, but some browser couldn't open with this kernel...
    
    No, this is still not fixed.
    • Clarissa Walker (amisapphire@cwcyrix.nsupdate.info)'s status on Monday, 23-Mar-2026 13:03:49 EDT Clarissa Walker

      in reply to
      It was to the point that I isolated NetworkManager due to early assumptions that it was to blame. In the end, it is a years-old kernel bug.
      
      Yes, my main laptop's wireless setup is entirely manual now, and I can easily revert this.