AEI (Hannover) issue tracker

2026-05-22

11:31 Warranty / Broken parts Vendor/misc (out of warranty) #3730 (Closed): a2911, a5311, a6111 lost connections to one GPU each

a2911:
[Tue Apr 21 11:12:44 2026] NVRM: Xid (PCI:0000:b2:00): 79, GPU has fallen off the bus.
[Tue Apr 21 11:12:44 ... Henning Fehrmann

2026-05-21

12:34 Atlas Bug #3729 (New): a reboot of the mk-storage node should yield a 20gbps speed setting for databond

/etc/network/interfaces.d/databond has the following content:
auto databond
iface databond inet dhcp
hwaddress a... Henning Fehrmann

2026-05-19

08:15 Warranty / Broken parts Vendor/megware #3728 (Resolved): Broken GPU 4 in g6139

GPU 4 shows Hardware failure in g6139 and need to be exchanged. Henning Fehrmann

2026-05-07

14:15 Atlas Hardware Repair #3723 (Resolved): a3108 doesn't boot up

This is the node where we added the V100. Could you take a loog, why this node comes up again? Henning Fehrmann

2026-02-18

15:57 Warranty / Broken parts Vendor/megware #3707: High ECC error rate in 2026-02

Get kernel messages on a7607:
```
[Wed Feb 18 15:44:07 2026] [Hardware Error]: cache level: L3/GEN, tx: GEN, mem-tx... Henning Fehrmann

2026-02-17

13:08 Atlas ToDo #3688: g6731 has eno2 connected to the data network

g6931 has shown some CRC error occurrences. It is currently not being used the MeerKAT pipeline. We might to check th... Henning Fehrmann

09:46 Warranty / Broken parts Vendor/megware #3707: High ECC error rate in 2026-02

These are occurrences in February 2026. I also saw CPU1 only affected. Henning Fehrmann

08:52 Warranty / Broken parts Vendor/megware #3707 (Resolved): High ECC error rate in 2026-02

Here is a list of nodes with very high ECC error rate in February 2026.
Do we have spare parts of should we start ca... Henning Fehrmann

2026-01-30

11:05 Atlas Hardware Repair #3694 (Resolved): g6731 and g8131 have network issues

g6731 and g8131 have a relatively high CRC error rate on the data network. After a reboot g8131 often negotiates ofte... Henning Fehrmann

2026-01-09

10:20 Atlas ToDo #3688 (New): g6731 has eno2 connected to the data network

Figure out why this is the case and possibly repair it. Henning Fehrmann