MAM59

I remember an event where cheap cables+cheap atlantic let to line errors and line resets. you would then note some dropping connections, usually client transfers are interrupted and the clients complain. But protocols like SMB do an automatic reconnect, so there is just a short delay and then transfers continue. Anyway in this situation it also happens that link loss also ends up into a full jam of the card. If flow control has stopped transfer and the line crashes then, the partner does not always get informed and the "resume transfer" packet is lost. The other box then is sitting around waiting until the cows come home. To see if this happens to you too do: wait until UNRAID is not reachable anymore Pull out the LAN Plug on UNRAID (or the switch side does not matter) Wait 5s Put in the LAN Cable again If you then can reach UNRAID again, you are a victim of the described problem. Either fix cables with correct ones, (and make sure the cable length is <30m!) or get a better card (and go for real fiber, these things happen only with twisted pair cables) Note: the cable stuff is tricky with 10G!!! There are still many many wrong cables on the market. For long times there was cable types CAT 6,7&8 available, but the plugs not. These cables were sold as "raw cable" (it is usually printed on them!), they DO NOT WORK! Read carefully!!! Note: the cable problem might also be on "the other side" of the connection. So take a look at the clients too!

He said, the stick (and its guid) stayed since 2020, but the guid that is stored at limetech mystically has changed ?!?!? very strange, besides contacting support and telling this unbelievable story, I have no idea what to do now.

If I have to guess, I would say, you have a serious problem with your DHCP server Dec 14 10:38:30 ATLANTIS dhcpcd-run-hooks: interface=br0, reason=IPV4LL, protocol=ipv4ll Dec 14 10:38:30 ATLANTIS update_services: remove queued job 3 Dec 14 10:38:30 ATLANTIS update_services: queue new job 4, wait for 20s Dec 14 10:38:30 ATLANTIS plugin-manager: executing inline script: /bin/bash '/tmp/inline6-dynamix.unraid.net.sh' Dec 14 10:38:30 ATLANTIS plugin-manager: executing inline script: /bin/bash '/tmp/inline8-dynamix.unraid.net.sh' Dec 14 10:38:31 ATLANTIS nmbd[2288]: [2025/12/14 10:38:31.345951, 0] ../../source3/libsmb/nmblib.c:929(send_udp) Dec 14 10:38:31 ATLANTIS nmbd[2288]: Packet send failed to 169.254.255.255(137) ERRNO=Network is unreachable Dec 14 10:38:31 ATLANTIS nmbd[2288]: [2025/12/14 10:38:31.346004, 0] ../../source3/nmbd/nmbd_packets.c:1639(retransmit_or_expire_response_records) Dec 14 10:38:31 ATLANTIS nmbd[2288]: retransmit_or_expire_response_records: Failed to resend packet id 11467 to IP 169.XXX.XXX.255 on subnet 169.254.111.85 Dec 14 10:38:31 ATLANTIS nmbd[2288]: [2025/12/14 10:38:31.346029, 0] ../../source3/libsmb/nmblib.c:929(send_udp) Dec 14 10:38:31 ATLANTIS nmbd[2288]: Packet send failed to 169.254.255.255(137) ERRNO=Network is unreachable UNRAID does not get an address and other config, so it is on its own and reverts to the 169.254.x range (which means: I HAVE NO NET AT ALL!). If this happens only SOMETIMES, it also could be the reason for your losses later on. Maybe on 1st boot UNRAID is able to fetch the data from the DHCP server and runs fine for a while. Then, the lease runs out hours or days (depending on the DHCP server's settings) and needs to be renewed. If the DHCP server is not reachable then, there will be an additional grace period in which UNRAID keeps on using the not-so-valid address. But after this (and still no answer from the server), the address and config will be deleted and switched back to 169.254.X. All these timeouts depend on the server's config, so I cannot guess how long it will take until you lose connection. so, what you have to do is to check the DHCP server and see why it does not answer.

you can safely ignore the rsyslogd errors. They are confusing but normal. They occur during boot phase if you have defined a remote host to store syslog on, but the network ist not yet up and running. The messages cannot be transmitted, so it writes down this errors. (rsyslog should be much brighter and know this and avoid the error messages)

The "normal" way to add a new disk would be just to: stop the array add the drive to the array list can't start the array because of the new drive but NOW there is a new button (actually 2) on the array start page! select the new disk from the list and check the checkmark box ("I really want this!") Done! The thing is, a new disk never has a proper filesystem and therefor needs to be formatted. Hint: check the disks settings before to set the filesystem type!!! Currently your disk has "auto" which means: "if there is already a filesystem on the drive determine which and try to mount it, else use the type from the global Disks Settings". Once acknowledged the disk will be formatted, mounted and the array started. (Sorry, no screenshot. these buttons only appear after you have added a new drive)

Yeah I know, but I'm afraid this is the only chance you have currently. Killing cats wont solve your problem 😁

Sounds like you have "bricked" it. This happened quite often in the early days of SSDs. The Firmwares were not really stable and could be killed "beyond recovery". Ask the vendor ("samsung"?) if there is a de-brick tool for this particular device. You man need to send it in (which obviously may give you additional costs, a new disk may be cheaper then). Maybe a firmware update is available and you can flash it yourself.

first of all: get rid of that bond! it is useless. Unplug all but one (eth0) cables. second: nobody will be able to help you unless you supply diagnostics here

Shit happens they say. I agree, its hard to build a cat-safe computer, we also have a bunch of cats, they examine everything :-) Anyway, you are somewhat wrong, "Linux" is NOT an operating system, its just a Kernel. Everything around it comes from other people/companies and yes, its a mess. These "distributions" differ alot, so what you read from Dist-X may not be true/valid/good for Dist-Y. UNRAID is its own distribution (based on others), which makes it unique. But I see hope for your insane SSD. Put it back into a windows system, open diskpart, select the drive and do "clean all" (this may take a long time, dont become nervous). Or use tools like "wipedisk" or something. At the end, the SSD will be free of all ext4 remains and can be freshly formatted and used in UNRAID. Use zfs or something the next time. Ext4 is not well supported in UNRAID yet (ok, I guess you have learned this part already now)

Erstmal die Protokolle der Docker lesen. Meist steht da schon im Klartext, was ihm nicht gefällt. Mangel beseitigen, fertig!

This will be hard and also uncessary. Just look for a free USB2.0 on your PC (there are surely some available, maybe internally? USB3.0 sticks will slow down and produce less heat (this is the 3.0 killer) In 2.0 mode the sticks usually will last much longer.

For real spinning drives you should leave the warning temps as they are on default (for SSDs they can be much higher). So you will have to think about how to improve your cooling. (BTW, its not uncommon that larger drives get hotter, there is more mass to spin around and so there is more energy needed which results in higher temperatures) Remember that drives that are stacked tightly even heat up each other. Maybe you can move around them, leaving empty slots between them? Or, sometimes it is also helpful if you reseat all drives by placing SSDs between the real drives (2.5" SSDs give more room for air flow in 3.5" cages) (HDD-SSD-HDD-SSD... and so on) There is no general solution, you have to see what works best for you.

likely you are a victim of the famous "2.5G needs Flow Control" issue. See if flow control is enabled on both ends and in the switch. if not, a new connection often pick the wrong sending window and it is not received by the other side. After a timeout it is retried within a different window. This can slow down a connection up to 75% (but usually it is only 50%)

Need a bit help with this. I'm using autofan control (and some other dynamix plugins), it work very well. But I find it a bit annoying how it spams the syslog with really not important messages like: Dec 9 11:04:02 f autofan: Highest disk temp is 40C, adjusting fan speed from: 129 (50% @ 703rpm) to: 150 (58% @ 817rpm) Dec 9 11:19:08 f autofan: Highest disk temp is 39C, adjusting fan speed from: 150 (58% @ 818rpm) to: 129 (50% @ 705rpm) Dec 9 11:23:14 f autofan: Highest disk temp is 39C, adjusting fan speed from: FULL (100% @ 1398rpm) to: 211 (82% @ 1175rpm) Dec 9 11:23:14 l3router syslogd: last message repeated 1 times Dec 9 11:28:19 f autofan: Highest disk temp is 30C, adjusting fan speed from: 211 (82% @ 1179rpm) to: OFF (0% @ 0rpm) Dec 9 11:28:19 l3router syslogd: last message repeated 1 times Dec 9 11:29:14 f autofan: Highest disk temp is 30C, adjusting fan speed from: 129 (50% @ 696rpm) to: OFF (0% @ 0rpm) Nice to know that it works, and, yeah, it might be reasonable if it has to go up to 100% with a fan (and maybe lower again later on), but I am not interested to see changes from 50 to 58% and back. Is there a way to silence it a bit so that other messages are more visible in the syslog???

delete the bond, unplug one card, reboot. Afterwards rethink what the bond should be good for? (under normal conditions NOBODY needs to have a bond and most of the people dont really know what it does and what it is good for (hint: it does NOT speedup anything!)) Obviously now you have a loop between the adapters, your switch is not playing along with the bond (or does not even know it should exist?)

your screenshot is not really readable and you have not addeds diagnostics. Unlikely that somebody can help you without these informations.

LOOOOOOOOOOOOOOOOOOOL Yeah Mikrotik Routers are great but not good for the untrained heart. You can do too many stupid settings. Normally ARP is a self contained and not configurable protocol, no need to twiddle around with it.

very unlikely! ARP is cached for 1hr only normally, down to 30s for non-answered requests. But of course, you can make it go crazy if you change cards faster than these timeouts The main reason is more likely that you were not cool enough and waited for the correct timeout (ARP ? address resolution protocol, aka a list which MAC currently represents one specific V4 IP address. For V6 something alike ist called NDP (neighbor discovery protocol). Does the same but works totally different.ARP is based on broadcasts whereas NDP is based on specific ICMPV6 packets to the F800:: address range)

could it be that your router/switch has a setting based on mac addresses in its firewall? If you have changed hardware this has changed too and no new card would match the old entry...

Rufus does a "cheap trick". It creates a very small FAT32 partition and puts a special EFI Bootloader on it. The remaining space is formatted with NTFS. The EFI Loader just mounts this NTFS partition and issues a 2nd boot from the EFI folder on the NTFS partition. The main reason for this is not to extend the FAT32 space, but to get rid of the 4Gb file size limit of FAT. The current windows versions (the .WIM files) are larger than 4Gb and Microsoft had to split them up into several files which makes the installation not really easier (the DVDs contain the large single files, if you create a stick, these are split up and cannot be edited afterwards anymore). So with a RUFUS formatted stick you can just copy over the DVD medium 1 to 1 and run setup from it. Of course, such a stick can also boot other OS's like UNRAID because the RUFUS Bootloader allows reading NTFS filesystems whereas BIOS/UEFI only need to support FAT.

Flow control (aka "Supported pause frame use" as called by the driver) must be enabled and used on both sides. Else it is not uncommon, that a freshly starting transmission picks the wrong slot and the packet is lost. It will be repeated afterwards and sooner or later they will get together again, but this takes some time and costs a lot of bandwith. 2,5G LAN is crippled. It is using 10G modulation scheme, but with 1 packet of data followed by 3 packets of pauses. Flow control is the method how they agree when it is time for data again. An unmanaged switch should simply pass on flow control information. But some of them are not done very well and drop the info. Chaos and Slowdowns will happen. But first check if it is enabled on both machines.

Works automatically, although UNRAID still shows it as "VFAT"

there is no need for <32G anymore. The restriction went away some Windows Versions earlier already. Now you can create an (ex)FAT Partition with more then 32Gb and modern PCs can boot from it (keep your BIOS recent!) But one thing from the old video still is important: put the stick into an USB2.0 port even if it is a 3.x one! This will slow it down and cool it down too. It will last much longer in 24/7 operations Boot speed will be faster slightly in 3.0 port, but you will regret it if the stick wears out quite fast. Boot Time is not really important, after the boot UNRAID runs from RAM, with full speed.

good 😘

not just retry, redo! :-) Again, you have to set the check boxes for the things you wanna keep (pools?) then hit APPLY, then hit DONE