TinyCore RedPill Loader (TCRP)

[haydibe]

 

18 hours ago, synoxpe said:

@haydibe you had inspired me to take the proxmox path 😀 and after a short lived bliss the environment has unfortunately become unusable.

Curious about your proxmox environment. Which version are you using? Can you please share the details?

I started with Proxmox 7.1-0 I think and kept them updated and am now on the latest 7.1-8. I do not have a subscription for any of my 3 nodes.  Two nodes use a Xeon E1275v5 and 32gb ram, the third uses a Intel i5-9500T and 32gb ram. All nodes use a Melanox 10gbps nic. I can't say that I experience any issues at all.  One of the 1275v2 Nodes has my main DSM VM with a passthough LSI adapter.   Before Proxmox, the nodes were operated as vSphere Cluster with a generated serial number - while this didn't bother me in the beginning it did after a while and I was looking for an open source solution. PVE uses KVM under the hood. May cloud solutions (AWS, every OpenStack Cloud) use it to provision and run their compute instances , so it can't be that bad.. can it?

 

Thus said, what details are you exactly looking for?

 

 

[Spagnoletto]

32 minutes ago, IG-88 said:

cpu resources would not be used by the kernel also if HT is involved (afaik) it uses normal and HT cores/threads evenly

6 cores + HT will be 4 real cores and 4 HT cores with 918+, ad "HT core" (or thread) will deliver 20-25% performance of a real core so if you do nothing you will end with ~5 core cpu performance compared to a 7.5 when using all resources (like using 3617/3622)

 you might mitigate by disabling HT in BIOS but the outcome depends on the cpu, so in case of 12 threads it would be 6 cores, slightly better then the 5 cores equivalent if HT would be still active (with 918+ and its 8 thread limit)

its not scientifically correct and might depend on what you are doing (virtualization with VMM might profit slightly when seeing 8 threads instead of 6 - but thats just guessing) and most of the time like in normal NAS (storage) operation there wont de any difference in performance because for just storage and software raid you dont need that much cpu resources

So if I Have two processor(server bare metal ) witch 6cores 12 threads each, on ds3615xs I have only one processor with 6 cores and 8 threads ?? Should I migrate to ds3617xs ??

Edited February 26, 2022 by Spagnoletto
change

[IG-88]

12 hours ago, Orphée said:

Actually it seems the Atom c3538 doesn't have igpu. So DVA3221 can't work with igpu... no?

thats true but to support object or face recognition in SS they would have to ffmpeg compiled in a way making use of nvidia's function's/api's and if that's the case photo station could be using it too

facial recognition is supported on non iGPU/QSV boxes but its not marked on what units its using hardware support be iGPU/GPU

https://kb.synology.com/en-us/DSM/tutorial/Which_Synology_NAS_models_support_the_facial_recognition_feature_on_Synology_Photos

 

i looked for some links between face recognition in "Photos" and DVA3221 but beside the link above i have not seen anything

as it would be a borderline case for DVA customers (mainly interested in real time survailance) i guess they would just use whats already working and add nvidia gpu support later, so if not mentioned i guess its not there

[IG-88]

13 hours ago, Spagnoletto said:

So if I Have two processor(server bare metal ) witch 6cores 12 threads each, on ds3615xs I have only one processor with 6 cores and 8 threads ?? Should I migrate to ds3617xs ??

it does not matter how many processors aka sockets are there, the limit is a kernel compile parameter and its just about threads, a thread can be one of a real core or just one "made up" by HT, if you look up on ark.intel you will see how they differ between core and threads

 

3615 hast the same limit as 918+ when it comes to threads and thats 8, in dsm 6.x only 3617 was better with 16 threads

with 7.0 3617 (broadwell) got bumped up to kernel 4.4 as broadwellnk (3622)  already was in 6.2 and also got 24 threads like broadwellnk already had in 6.2

3615 and 918+ are still the same in 7.0

there are experiments going on with FS6400, that comes with 64 threads/cpu's but for now 24 with 7.0 is max we have

only advantage on 918+ for now is intel quick sync support (i915 driver) and i plan to test this weekend if a additional i915 driver is possible with kernel 4.4 units like 3622 (depends on the the compile settings of the kernel, might not work as it did not in 6.2 where it was on kernel 3.10, that still had i915 support even if it was pretty old)

 

Edited February 27, 2022 by IG-88

[synoxpe]

3 hours ago, haydibe said:

 

I started with Proxmox 7.1-0 I think and kept them updated and am now on the latest 7.1-8. I do not have a subscription for any of my 3 nodes.  Two nodes use a Xeon E1275v5 and 32gb ram, the third uses a Intel i5-9500T and 32gb ram. All nodes use a Melanox 10gbps nic. I can't say that I experience any issues at all.  One of the 1275v2 Nodes has my main DSM VM with a passthough LSI adapter.   Before Proxmox, the nodes were operated as vSphere Cluster with a generated serial number - while this didn't bother me in the beginning it did after a while and I was looking for an open source solution. PVE uses KVM under the hood. May cloud solutions (AWS, every OpenStack Cloud) use it to provision and run their compute instances , so it can't be that bad.. can it?

 

Thus said, what details are you exactly looking for?

 

 

Thanks for responding @haydibe. I’m running into the Intel Nuc e1000e hardware hang issue which has started to affect the stability of the whole PVE environment all of a sudden. I was using kernel 5.13.9-4. Rolled back to 5.13.9-3 and -2 as well without respite. Seems like I need to reinstall proxmox 7 from the downloaded iso if the this helps. Else I need to turn off the checksum offloading as explained in some workarounds. 

[Orphée]

Hi,

 

So I successfully migrated my HP Gen8 from ESXi to Proxmox VE after fighting with relax inter rmrr patch to enable passthrough.

 

It currently works, but where ESXi have the issue with first disk /dev/sda always reserved and LSI HBA IT card beginning at /dev/sdb (SataPortMap=18 / DiskIdxMap=1000)

 

With Proxmox there is 2 virtual sata controller (I don't know where they are set), the LSI HBA IT card begins at /dev/sdc. I must for this set : SataPortMap=118.

I tried DiskIdxMap=100A00 but it does nothing.

 

So same as ESXi, with LSI card passthrough the DiskIdxMap settings is ignored.

 

I don't know yet if there is a real benefit to change from ESXi to Proxmox, but at least for the proof of concept I did it.

[brickrat]

On 2/25/2022 at 3:55 PM, Orphée said:

It has been explain multiple times already.

 

SataPortMap define the number of ports each SATA controller contains. set on 1 digit.

58 means : First SATA controller has 5 ports (5 pluggable SATA disks), Second SATA controller has 8 ports (8 pluggable SATA disks)

 

DiskIdxMap define the starting position of disk mapping for each controller, set on 2 digits.

00 = /dev/sda

01 = /dev/sdb

etc...

 

so 0A00 means :

- First 2 digits are 0A : is the 11th position : /dev/sdk

- 2nd 2 digits are 00 : it is the first potition : /dev/sda

 

So first controller disks will start from sdk to sdo

and 2nd controller disk will start from sda to sdh

 

Depending on how many SATA controller you have on your motherboard (are they all on motherboard, or do you have a PCI-card, or nmve... etc...) the number can be different from one to another configuration.

 

Whereas on ESXi you define the SATA controller numbers (like SATA0:0 for loader, and SATA1:x for DATA disks)

usually using the above settings, you are able to set SataPortMap = 18 and DiskIdxMap to something like 1000

As you only have the loader set on SATA0 you can have first digit set to only 1, where 2nd number will be 8 to be able to have 8 (virtual or RDM) disks plugged on SATA controller 1.

Hi

I am still struggling with this.

 

I have a MSI desktop board, so this should have 1 controller with 4 SATA ports

Does not matter what options I use here I get either not disks found or some sata ports disabled.

 

Can someone please assist and tell me what I have to enter in the config file.

[yanjun]

6 hours ago, IG-88 said:

it does not matter how many processors aka sockets are there, the limit is a kernel compile parameter and its just about threads, a thread can be one of a real core or just one "made up" by HT, if you look up on ark.intel you will see how they differ between core and threads

 

3515 hast the same limit as 918+ when it comes to threads and thats 8, in dsm 6.x only 3617 was better with 16 threads

with 7.0 3617 (broadwell) got bumped up to kernel 4.4 as broadwellnk (3622)  already was in 6.2 and also got 24 threads like broadwellnk already had in 6.2

3615 and 918+ are still the same in 7.0

there are experiments going on with FS6400, that comes with 64 threads/cpu's but for now 24 with 7.0 is max we have

only advantage on 918+ for now is intel quick sync support (i915 driver) and i plan to test this weekend if a additional i915 driver is possible with kernel 4.4 units like 3622 (depends on the the compile settings of the kernel, might not work as it did not in 6.2 where it was on kernel 3.10, that still had i915 support even if it was pretty old)

According to the research results of @jumkey, any model with the SYNO_PORT_MAPPING_V2 option turned on in the kernel will have a model.dtb, which describes the pci bus location, hard disk number and other information of the device of this model. Once it does not match, it will be installed during installation. It says that the hard disk is not inserted. The case of DS920 (geminilake) is relatively simple, it only needs to detect the bus position of the hard disk, while the FS6400 is much more complicated, it will detect two necessary hba cards. So maybe we need to emulate a PCI card for the FS6400? I have no idea.

 

The good news, though, is that you can even add 60 hard drives to the FS6400.

 

FS6400.dts

Releated Error.

 

 

Edited February 26, 2022 by yanjun

[yanjun]

On 2/19/2022 at 10:01 PM, Peter Suh said:

 

There was a similar phenomenon in my DS3622xs+ SHR.
SHR was first configured with four SATA HDD.
Unexpectedly, I needed a day of restructuring.
My 8GB of RAM was using almost 95% of the resources.
So, I deleted the storage pool and volume and made SHR again with only one disk, and did not attempt to reconstruct it.
If there is another attempt to reconstruct with two or more disks and another lack of RAM resources occurs, it plans to report it to @pocopico

 

On 2/20/2022 at 6:28 AM, Peter Suh said:

 

The first shr-generated multi-disk combination I tried was a test of 0.6 TB + 2 TB + 2 TB + 3 TB. After the RAM issue occurred, the storage pool and volume were deleted, and abnormal restart occurred in this process. Since this restart, the storage pool has been deleted and there have been no other errors. I don't know if I know exactly, but the mpt3sas driver seems to have been updated from version 22 to 41, so I'm rebuilding the loader and watching for 18 hours after creating it with SHR 2TB+ 2TB+3TB. No other abnormal symptoms have been seen yet.

 

On 2/20/2022 at 10:01 AM, Peter Suh said:

 

I'm sure some of you already know,

For repeated SHR or RAID volume configuration tests,

Share tips on stopping volume optimization.

 

SSH>

sudo -i

cat /proc/mdstat (Checking the disk number for optimization.)

echo frozen > /sys/block/<md with number>/md/sync_action

After this process, the storage pool and volume can be removed immediately.

 

In the last RAM issue, symptoms were witnessed shortly after SHR optimization.
SHR optimization will continue to try and report any problems.

 

II also encountered RAM (memory) problems. When I used qbittorrent to download files, I encountered dozens of restart problems. I used prxomox and allocated 40G of memory for DSM. Later, I found that it basically happens when the memory is full. reboot. The configuration of 3622xs+ on memory seems to be support_memory_compatibility=yes written in synoinfo.conf. If anyone has encountered a similar problem, I hope you can give feedback on the usage scenario so as to further locate the problem.

[haydibe]

4 hours ago, Orphée said:

With Proxmox there is 2 virtual sata controller (I don't know where they are set), the LSI HBA IT card begins at /dev/sdc. I must for this set : SataPortMap=118.

I tried DiskIdxMap=100A00 but it does nothing.

 

So same as ESXi, with LSI card passthrough the DiskIdxMap settings is ignored.

Odd how a passthrough adapter does that. With a Q35 type PVE machine there are always 2 sata controllers connected, while all virtual disks will be attached to the second. I have no idea why the first even exists. You could use remap to straighten out the drive order, I remember it beeing discussed in this thread. the remap string requires escaping and can be used to straighten out the driver ordering.

[Orphée]

18 minutes ago, haydibe said:

Odd how a passthrough adapter does that. With a Q35 type PVE machine there are always 2 sata controllers connected, while all virtual disks will be attached to the second. I have no idea why the first even exists. You could use remap to straighten out the drive order, I remember it beeing discussed in this thread. the remap string requires escaping and can be used to straighten out the driver ordering.

found it there :

 

Quote

"sata_remap": "24\\>2:25\\>3:26\\>10:27\\>11:28\\>12:29\\>13"

 

for me it would be something like :

"sata_remap": "3\\>1:4\\>2:5\\>3:6\\>4:1\\>8:2\\>9"

starting from 1, so -1 if starting from 0

 

right ?

[haydibe]

Great that you found the post, Orphée! I still find it odd that > even needs to be escapped with \\, even though it shouldn't be necessary in a json value.

It should use zero offset as well, but tbh I am not sure if the offset starts at 0 or 1.

 

 

[Peter Suh]

9 hours ago, yanjun said:

 

 

 

II also encountered RAM (memory) problems. When I used qbittorrent to download files, I encountered dozens of restart problems. I used prxomox and allocated 40G of memory for DSM. Later, I found that it basically happens when the memory is full. reboot. The configuration of 3622xs+ on memory seems to be support_memory_compatibility=yes written in synoinfo.conf. If anyone has encountered a similar problem, I hope you can give feedback on the usage scenario so as to further locate the problem.

 

I have used esxi once, but Proxmox vm has no experience. I would appreciate it if someone familiar with Proxmox could check and help.

 

[synoxpe]

What would be a good starting point to troubleshoot an improper shutdown of an instance [ds918+ running on proxmox]?

If I ssh in to DSM, I only can see dmesg messages from when the boot up started and not earlier, so it’s not possible to see a possible root cause.

@IG-88 @Orphée

[IG-88]

5 minutes ago, synoxpe said:

What would be a good starting point to troubleshoot an improper shutdown of an instance [ds918+ running on proxmox]?

If I ssh in to DSM, I only can see dmesg messages from when the boot up started and not earlier, so it’s not possible to see a possible root cause.

@IG-88 @Orphée

add a serial com port and connect to it with a terminal, can be seen in this configuration, connection parameters can be seen in grub.cfg

https://xpenology.com/forum/topic/7387-tutorial-dsm-6x-on-proxmox/?do=findComment&comment=147159

 

i use that on baremetal and virtualbox vm's

[synoxpe]

53 minutes ago, IG-88 said:

add a serial com port and connect to it with a terminal, can be seen in this configuration, connection parameters can be seen in grub.cfg

https://xpenology.com/forum/topic/7387-tutorial-dsm-6x-on-proxmox/?do=findComment&comment=147159

 

i use that on baremetal and virtualbox vm's

I’m using proxmox 7, DSM 7 using the pocopico tinycore loader 0.4.4. Is the link that you pointed still applicable?

[Orphée]

What do you mean by improper shutdown ? while DSM starting ? after couple of minutes ? when shutting it down ?

[IG-88]

30 minutes ago, synoxpe said:

I’m using proxmox 7, DSM 7 using the pocopico tinycore loader 0.4.4. Is the link that you pointed still applicable?

sure, its a config for proxmox, you can add the serial com port to your proxmox vm as in the link and thats it, ready to use

dsm is meant to be used as a headless system, all important output is sent to the serial console when booting

Edited February 27, 2022 by IG-88

[AmiroGarry]

is it possible to create a bootable UEFI image? With the new CPUs, CSM is no longer possible without an external graphics card. Thus, an image is not possible to create. The current image only works with CSM active. This is no longer possible with the new boards. Since then the internal graphics card from the Intel CPU no longer works.

[IG-88]

17 minutes ago, AmiroGarry said:

is it possible to create a bootable UEFI image? With the new CPUs, CSM is no longer possible without an external graphics card. Thus, an image is not possible to create. The current image only works with CSM active. This is no longer possible with the new boards. Since then the internal graphics card from the Intel CPU no longer works.

i use a hp deskpro desktop for testing and even without CSM disabled the MBR bootable usb does its job and boots in uefi boot source

i guess by "uefi image" you mean a GPT bootable media?

there are tools where you can convert mbr/gpt, like "MiniTool Partitinon Wizard Free", maybe try that one to convert the usb you have into gpt

 

Edited February 27, 2022 by IG-88

[AmiroGarry]

on all my computers if UEFI only is set without CSM active, no boot with the Tinycore loader works. Have Tinycore separately downloaded is on the image. Tinycore gives the larger image here, UEFI is properly supported. Once the image is compiled, the UEFI boot works.

[synoxpe]

1 hour ago, Orphée said:

What do you mean by improper shutdown ? while DSM starting ? after couple of minutes ? when shutting it down ?

DSM is starting fine but crashes spontaneously after running for some time, which ranges from 4 mins to 9 hours. Sometimes I see the confounded e1000e hardware hang messages on proxmox syslog, sometimes not.
But the DSM crash happens inevitably and all I get is an email from DSM (which I’ve configured) warning about an “improper shutdown”. 

The crash is so frequent sometimes that DSM automatically goes to data scrubbing mode to check data integrity, and the crash happens yet again after sometime.

As of now I’ve shut everything down to prevent potential data loss.

Any pointers are wholeheartedly welcome!!

[synoxpe]

2 minutes ago, AmiroGarry said:

on all my computers if UEFI only is set without CSM active, no boot with the Tinycore loader works. Have Tinycore separately downloaded is on the image. Tinycore gives the larger image here, UEFI is properly supported. Once the image is compiled, the UEFI boot works.

UEFI boot worked perfectly with v0.4.5 image on my Intel NUC, I had to turn off secure boot though.

[synoxpe]

1 hour ago, IG-88 said:

sure, its a config for proxmox, you can add the serial com port to your proxmox vm as in the link and thats it, ready to use

dsm is meant to be used as a headless system, all important output is sent to the serial console when booting

Awesome man. Will give it a try when I’m back into the arena!

[AmiroGarry]

I've only tested 0.44 so far. Is UEFI fixed in 0.45?