Search results

disable NUMA

So adding a rootdelay kernel parameter helps grub, the pre-kernel environment?

@mbaldini: this machine stops at grub, before inird's busybox.

Latest 4.13 (build Nov 6) from pve-test randomly starts a kworker process that hogs the CPU like crazy and the performance is reduced to a crawl.

Bots can't read statistics :)

I think ZFS does make a difference between non-zero and zero (actually compressible data), but you need compression enabled for that. The highly compressible data does not have a block allocated at all, being embedded in block pointer. https://illumos.org/issues/4757 zpool upgrade -v ...

I am running a gen8 microserver using Proxmox 3 and then 4 for two years now. CPU is E3-1265L V2. In the past I've had VMs on Windows 7 & Server 2012R2. Never had an issue.

I assume you know that (almost) everything at Google runs in containers, right?

For home storage I'm using some ZFS pools (mirror for important data, striped for other) and export them through FTP, AFP, CIFS, WebDAV using containers (you can mount host folders inside a container).

You should put a question, to get an answer. Anyway, the reason why you are not able to see them is because they are not mounted. They are not mounted because the folder structure is already there when zfs mount runs. Search for this, there were some solutions.

cat /proc/net/iet/volume Nevermind, that's for target, sorry. ls -l /dev/disk/by-path Example: $ ls -l /dev/disk/by-path/ip*-lun-0 lrwxrwxrwx 1 root root 9 Dec 15 17:37 /dev/disk/by-path/ip-10.129.30.41:3261-iscsi-iqn.cn02:brick1-lun-0 -> ../../sdm lrwxrwxrwx 1 root root 9 Dec 15 17:40...

That would be "kvm_amd"?

Yes.

These are kernel values, LXC containers do not have a kernel, so you need access to host kernel. Either a privileged container, or, if you say you want these in all containers, just set them on host.

Node IP != Cluster IP in all my books.

The recommended installation method for Kubernetes nodes is using VMs, therefore Proxmox KVM support is fine. If you really need an UI, you have kubernetes-dashboard (with heapster for graphs). The best way to interact with Kubernetes is through programatic means (entities descriptors, CI push...

Nope. They are not. Async writes are buffered in memory and that is an OS thing, not controller. Maybe the controller driver signals to the kernel that can send all writes in sync or you mount your filesystems in sync, then you will get all your writes directly in controller cache.

True. That is why we use slog devices.

Maybe few SSDs for slog. The controller cache is not SATA/SAS bound, it is PCIex, so many many times faster. To be honest, I don't think you will feel a difference in real life.

More or less. The write bandwidth of the SSD will be smaller than the write bandwidth of a 1 GB cache RAM for sure.