12:03:47 So my NAS got ipV6 address, however after 1-2 hours the ipv6 address disapears. Restarting the ndp service restores the ipv6 adress: https://i.kek.sh/7uN27YLGMC0.png But then it disappears again. What have I missed during the setup, how can I set up the NAS to keep the IPV6 adress? I am using SLAAC here. 12:13:10 root@omnios:~# cat /etc/inet/ndpd.conf 12:13:11 ifdefault StatelessAddrConf on 12:13:11 ifdefault StatefulAddrConf off 12:13:11 ifdefault TmpAddrsEnabled true 12:13:13 ifdefault TmpPreferredLifetime 600 12:13:15 ifdefault TmpValidLifetime 600 12:15:05 Might be rhe tmpaddrsenabled 12:16:05 I had this before I have added the tmpaddr too. 12:16:10 root@omnios:~# cat /etc/hostname6.e1000g0 12:16:11 token ::10/64 12:16:20 I also have this. 12:16:59 I didn't had IPV6 connectivvity during the install, so maybe I missed some important setting / file / service. 12:31:13 nas is omnios? 12:38:17 yes 12:40:34 does ipadm show address objects for v6? 12:41:05 have you see nthe screenshot I posted? Does that contain the info you was asking for? 12:41:16 like: vioif0/v6 addrconf ok fe80::8:20ff:fe4d:a12a%vioif0/10 12:41:17 https://i.kek.sh/7uN27YLGMC0.png 12:41:46 I have censored it because it is a global ipv6 adress 12:42:09 ok, objects are there. 12:42:29 this is physical host with direct connection or vm? 12:44:07 this is a physical host 12:45:31 essentially, if your setup is using stateless addrconf, then the router is supposed to send periodically out prefix announcements, if those announcements stop, then the prefix will expire and the address is dropped from interface. With dhcpv6, you get stateful config and address (with prefix) is coming from dhcp server and lease lifetime settings apply. 12:47:03 But as I have mentioned, restarting the service makes the ipv6 address show up immediately. So I think the router announces the prefix just fine, just omnios ignores it and lets the address expire. 12:47:27 in case of stateless setup, it is possible your ISP is not sending prefix announcements (then the prefix will expire in your router and it will stop announcements in your lan), or your router is buggy. 12:48:09 you should be able to monitor the announcements with snoop/tcpdump/tshark 12:49:11 I had similar issue with my netgear orbi, and at some point in time it got "fixed", I'm not even sure if the cause was at ISP or in my router... 12:49:16 I understand, but none of the other clients have any issue with the router. So I suspect the issue is in omnios or in its config. 12:50:12 I have created the ndpd.conf and the ipv6 address did not dissapeared in the last 30 min. I will monitor it further and hope it will work now. 12:50:20 oh, rhight, ndpd has option to switch on the debug printout 12:50:35 the ndpd log shows me the following: 12:50:49 Mar 16 13:10:37 /etc/inet/ndpd.conf line 6: TmpDesyncFactor (600) is greater than TmpPreferredLifetime (120) 12:50:52 Mar 16 13:10:37 /etc/inet/ndpd.conf line 7: TmpPreferredLifetime (86400) is greater than TmpValidLifetime (120) 12:51:29 86400 is 24h. I suspect omniOS simply lets the address expire. So i have reordered the entries in the ndpd.conf like this: 12:51:47 ifdefault TmpPreferredLifetime 600 12:51:47 ifdefault TmpValidLifetime 600 12:52:23 since then no warning. MAybe the default 86400 sec alue is simply too long. 12:54:11 but this config options are for the temp addresses, so it should not impact the global ipv6 address. 12:54:19 yep. 12:54:36 no idea then :S 12:55:33 well, you can run in.ndpd -d (or -D XXX where XXX is value you can find from source), this will stop turning into daemon, so you need to disable smf service and start it manually 12:56:01 -d does enable full debug, so it will spam some... 12:57:05 but my symptoms were quite similar -- phones and macos had ipv6, illumos got them expired 13:01:33 I was already suspecting if its related to missing ND_OPT_ROUTE_INFO feature, but then it was "magically" fixed:D 13:08:01 ipv6 adress just disapeared. 13:08:07 Lets try with the -d 13:09:29 strange, but I can1t sem to find the in.ndpd executable, it is not in $PATH 13:09:58 /usr/lib/inet/in.ndpd 13:11:36 yep 13:25:33 I have disabled the tmpddrs to reduce the noise and got his: https://paste.omnios.org/?259a1b79231d6983#CmUXyjYmLtM7kmbZkQrdU6aGV7XbZPBt3zRG42hmrrBr 13:25:43 The router is running up-to-date openwrt. 13:44:45 you can trace only prefix with ndpd -D 0x0008, constants are in https://src.illumos.org/source/xref/illumos-gate/usr/src/cmd/cmd-inet/usr.lib/in.ndpd/defs.h?r=176dc728&mo=2271&fi=90#86 14:08:03 It says: "Address removed due to timeout " 14:08:13 whole log: https://paste.omnios.org/?bda5cff789176b29#9pavKsqFNQyMKUYiFhvr5UJVMP5qRyXr5yCBKStCvNYk 14:43:37 I have buped the "IPv6 Valid Prefix Lifetime" and "IPv6 Preferred Prefix Lifetime" to 10 years. 15:56:31 szilard: router should be periodically multicasting a prefix advertisement to ff02::1 ; if ndpd doesn't see the advertisement it will time out the prefix and the address will go away. 21:11:40 And... second time my box has just randomly hung, then self rebooted with no crashdump or other clue as to why. First on 151054ad, this time on 151054am. 'feels' like a storage IO hang as it'll ping but not complete an ssh attempt before it reboots? 21:11:54 Of course it's remote so no idea what was on screen when it tanked. 21:13:46 Sometimes I miss the old hard-copy console. 21:14:00 Yeah.... would be so nice right about now. 21:14:13 Save for the hour drive to get to it... 21:16:36 RS-232 console server with log .... 21:24:26 smartctl looks healthy across the board 21:34:10 nvme slogs show healthy 21:52:17 Ok, https://www.illumos.org/issues/8391 matches up pretty well to what I've observed, short of seeing the console output. 21:52:18 → BUG 8391: mr_sas controller lockup on Dell H330 (New) 21:52:29 https://illumos.topicbox.com/groups/developer/Td11518c45ba141cc as well 21:55:41 if you can, replace it with a HBA330 21:59:08 Or, toss it given the mandate is no more local anything... blech. This will be a fun chat tomorrow. 22:04:56 Can I convince work to spend $65 to fix it short term so I'm not down storage while migrating to AWS? 22:06:41 i paid something like $25 for my hba330 :) 22:08:27 The trick is convincing for ANY spend on HW is now stupid difficult. 22:09:50 Kurlon: is the hardware otherwise good? convince them to let you take it for $homelab. 22:10:45 R640 with 512GB RAM, I suspect they're going to try and sell it despite it being EOL, but I'm hoping otherwise. :D 22:11:10 EOL doesn't mean it can't be resold... just not for as much $$$. 22:11:30 * nomad doesn't deal with dell so has no idea if that's good hardware or not. 22:12:04 It's OK for it's age, loud as heck and not exactly skimping on power draw for homelab use. 22:12:41 so.. classic :) 22:12:42 Other than this problem, it's been rock solid for the 7, 8? years we've had them. 22:14:21 Take the HW, find a matching generation bare workstation to transpose the ram / cpu(s) into, add big ol video card... battle station status: Excellent 22:15:37 If the lights in the block don't dim when you hit power, it's not big enough.