Arch Linux

justm3

Member

Registered: 2019-11-12

Posts: 56

Chrome and RAID10

I am dealing with a weird situation and no clue to go anywhere.

Running a fully updated box with KDE Plasma on top, RAID10 (3 disks), 5.4 kernel-lts, 3GB of ram and an equal sized swapfile, an old nvidia with 390 driver. The system runs smooth with no issues until I load google-chrome.
A couple of tabs later (random, no specific tabs) the system stops responding, hangs(even the mouse does not move-keyboard is not responding) and raid synchronization starts at a very-very-very low pace.

Nothing responds, I mean nothing.
It would take half an hour for chrome to close, another half hour for cli to respond to mdadm -D and show me that it syncs.

The issue does not seem to repeat itself if I load firefox (can't say 100% because I am not really using firefox that much). I am posting with firefox right now and the box is on for the last couple of hours-as a reference.
The weird thing is that I get no journalctl messages. Or any other messages that I've searched for (-b for example).

I ran a scrubbing crontab which I deleted just in case something was wrong from my side. Same scenario.

Another thing that puzzles me is that every time this happens and I ran cat /sys/block/md0/md/mismatch_cnt the bad blocks increase.
But that is not the case with bad blocks after scrubbing. Hence I tend to believe there is nothing wrong with the spinners or the array.

I can't think of what would be appropriate to supply as info.
Any ideas what to look for? I mean other than drop chrome (need the cast).
Thanks in advance.

frostschutz

Member

Registered: 2013-11-15

Posts: 1,652

Re: Chrome and RAID10

raid doesn't resync for no reason, when that happens the kernel is usually quite verbose about it. check dmesg / journalctl when it happens.

Slithery

Administrator

From: Norfolk, UK

Registered: 2013-12-01

Posts: 5,776

Re: Chrome and RAID10

How the hell are you doing raid 10 with only 3 drives?

---

No, it didn't "fix" anything. It just shifted the brokeness one space to the right. - jasonwryan
Closing -- for deletion; Banning -- for muppetry. - jasonwryan

aur - dotfiles

justm3

Member

Registered: 2019-11-12

Posts: 56

Re: Chrome and RAID10

Tks for the replies.

@frostschutz
I will try and if I manage to get it spit-out info will see if I can copy it.

@Slithery
I guess very straight forward.
mdadm --create --level=10 --metadata=1.2 --chunk=256 --raid-devices=3 --layout=f2 /dev/md/md0 /dev/sda1 /dev/sdb1 /dev/sdc1
Ok I know that technically it's not r10 but rather what is called r1E but in real life you get the benefits of both worlds (speed because this is an old box and minimum resilience in case one disk fails)
Why? Do you think I do something wrong?
It really works nice and fast (these 3 disks are old 7400rpm spinners). The whole system is pretty responsive (with the exception of the issue on this thread)

justm3

Member

Registered: 2019-11-12

Posts: 56

Re: Chrome and RAID10

An update.
The issue repeats itself with random apps not only chrome (which is open at all times).

It was never-time-ending situation to get a response from cli and check the memory. Once I got a view, it was practically flooded, hence no response from the system.

Hard rebooted and decided to move away from /swapfile and installed systemd-swap with zswap_enabled=1 and swapfc_enabled=1. I also increased swapfc_chunk_size to 1024M and allowed 8x.

Seems to work fine. Opened as many apps as I figured, chrome with multiple tabs (twitter feeds, youtube streams, kodi streams can't remember all).
The system slowed down only at the point where swapfc increased (doubled). Other than that it was ok.