That looking into a folder path length issue I found several hundred folders that look as if someone wrote a novel in the file name/path.
what did you learn today? (part 2)
- Thread starter continuum
- Start date
I only ever notice such things when my backups fail because of them. One time, I had some sort of recursively linked folder that was infinite. No idea how that happened, and I think I had to delete it via a linux boot cd.
That is how I discovered this mess. MoveIT failed on a job due to error in path
We’ve got a customer who insists on using the customer’s name and address and customer number when creating job folders because that’s the only way everyone can easily find the correct folder. Or so the PMs claim.
moosemaimer
Ars Scholae Palatinae
I had someone complaining that when they tried to save changes to a file they'd created, it was telling them they didn't have permission. I eventually traced it down to a bug in LibreOffice (dating back to OpenOffice) where if you had a '#' in the path, on a Samba share, you could create but not edit files.
Wow, nice troubleshooting. Ye gods that would be difficult to track down.
Was it intended as a feature of some kind?
moosemaimer
Ars Scholae Palatinae
I'm betting it was a failure to escape something correctly in the routine to save over an existing file.
Plz go upvote my OS request addition for this link shortener - https://github.com/shlinkio/shlink/discussions/2101#discussioncomment-11319142
...that someone came in and manually shut off my field server UPS without a proper shutdown. Maybe the Netapp survived but the Cisco Nexus hangs at the POST copyright. Yay.
More votes for the never cisco camp. Is there even a way to turn it off? All network gear I've ever seen says just yank power to turn them off.
Palo Alto has the extra shoot yourself in the foot option of powering off the management plane and the only way to get the box to restart is physically powercycle it. It warns you of this but I'm not sure why it's even an option.
Palo Alto has the extra shoot yourself in the foot option of powering off the management plane and the only way to get the box to restart is physically powercycle it. It warns you of this but I'm not sure why it's even an option.
that apparently we have a bunch of devices that use a actual PC name as a NTP server, instead of IP.
FWIW with our Juniper gear we were taught to use request system power-off or request system halt to gracefully power down.
That's because a whole series of their switches were a dumpster fire that ran the file systems in read write mode and not from a overlay/ramdisk and had a high chance of breaking the filesystem if not unmounted. Stupid things, (EX3000?)
Last edited:
Using DNS would be understandable if it was ntp.wherever.local or something like that
I agree and will likely be the solution here. This is yet another case of info not fully documented or communicated.
At my last job we anycast'd it to 10.123.123.123 and 26xx:11x:1xxx:123::123
<insert longwinded bitch session about fucking Ubiquiti and their fucking Edgerouters and their shitty IPSEC implementation>
Fucking Mikrotiks at least worked properly as dumb VPN appliances.
Fucking Mikrotiks at least worked properly as dumb VPN appliances.
They're two sides of the same coin, horrible in their own ways but man They're cheap.
Too many people treat them as the best thing since sliced bread.
We tried out some CCR2004s, ran into a ton of actual issues on top of the annoyances with their ecosystem and was basically told by the community that I should have known better than to get a product that wasn't a few years old and crazy shit like "you made too many configuration changes and the flash is corrupt but you can't see that, so you need to net install it to reset the flash partitions"
Yeah no thanks, they were replaced with a pair of Arista 7280R3's. How many times do you think those have not saved a config properly after a write mem?
(Yes I understand comparing a CCR to a 7280 is like comparing a Fiat 500 to a Taycan)
Too many people treat them as the best thing since sliced bread.
We tried out some CCR2004s, ran into a ton of actual issues on top of the annoyances with their ecosystem and was basically told by the community that I should have known better than to get a product that wasn't a few years old and crazy shit like "you made too many configuration changes and the flash is corrupt but you can't see that, so you need to net install it to reset the flash partitions"
Yeah no thanks, they were replaced with a pair of Arista 7280R3's. How many times do you think those have not saved a config properly after a write mem?
(Yes I understand comparing a CCR to a 7280 is like comparing a Fiat 500 to a Taycan)
We had decent luck with our CCRs for a while but the Juniper ACXes that replaced them were considerable improvements.
Odd, those are usually fairly resilient since anything important is kept in battery/supercap backed caches.
Netapp's own site won't let me see the fucking KB article telling me how to restart the applications without an account and their account registration apparently takes a day to kick in. Fucked up thing is that I've had a Netapp account before but we changed email back in 2021 so new account it is.
Put the controllers in loader prompt and disabled autoboot to charge those batteries juuuuust in case.
TIL... I have Monoprice mice... At least the handful of cheap ones I got as disposable spares for people. I think I'd rather not have a mouse. It works ok on a sheet of paper, but not directly on any desk... time to find a mouse pad
We generally use DNS to ntp(x), but they all map to a .123 address.
I was wondering, "why .123 for NTP?", and then I remembered the port number.
Still probably wouldn't put a server there, my reflex is to keep those low and high to make room for a nice big DHCP block in the middle. (edit: well, servers low, network gear high, to be more precise.)
Still probably wouldn't put a server there, my reflex is to keep those low and high to make room for a nice big DHCP block in the middle. (edit: well, servers low, network gear high, to be more precise.)
At scale servers shouldn't be in the same subnet as their management interfaces. And as I said, we were big/complex enough to do internal anycast for select services, so we had no issues injecting 10.123.123.123/32 into the routing tables.