Aaron's blog

Halloween candy

We gave out fruit snacks, which came in a box of 160 packages. Each one contains a stingy 6 gummies.  When I counted this morning, 90 remained. We ate four last night, leaving 66 given to trick-or-treaters.

I ate two this morning while counting.

Mandan house trees

Pickup loaded with trees

For posterity, here is a description of the trees that we planted in September 2021.  All were purchased from Prairie View.  I'll say they were fine but I'll check out somewhere else next time.  We went with fairly large trees and I think it was a worthwhile expense do to so.  They were all extensively watered for the first fall and then two summers.  I did almost no watering in 2024, which was helpfully a pretty wet year.

Boulevard Linden

We planted two of these in the boulevard at $327 each. These seem to be slow growers but I'm sure someday they will be mighty trees!  Just now in 2024 we finally see significant growth on the top of the north tree (more sun on that one since the south tree is shaded--my theory).  Info 

 

Medora Juniper

A sort of privacy screen is formed by these in the side yard.  $96.50 each.  These have been growing at a great pace and staying very shapely.  Water/ice falling from the porch onto one has not really harmed it so the hardiness is impressive.  Info

 

Dwarf Korean Lilac

Three of these serve as show-pieces at the very front of the yard.  I have been able to keep them trimmed in nice ball shapes.  Ideally these would be just big enough that they go just up to the sidewalk but not over the edge.  They have several years before that will be a problem.  These are grafted trees, $218 each.  The flowers only seem to last two days.  Info

 

Northern Empress Elm

This tree is in the middle of the front yard and has really started to nicely shade the porch area.  It has grown rapidly and has an impressive trunk.  Rated to grow 28 feet tall and 24 feet wide, it should not overwhelm the front yard.  It grows a ton of downwards drooping garbage branches so I get lots of practice pruning.  Apparently this was first available in 2021, the year we got it.  $312 for the pleasure!  Info

 

Hot Wings Tatarian Maple

I thought for sure that the leaves would be the hot-wings-colored part of this, but I see now that it is the seeds.  That does make sense, since the seeds (samaras, aka helicopters) are like wings, and that is what they say turns red.  $186 for this one.  It had 6 main shoots going straight up in a bundle for the first two years, and they were very annoying in the way that they tangled with each other, but now they have matured out to do their own thing and the tree looks good.  This one is in the side yard.  I imagine that people think it is dumb to put a maple tree right between two kind of close houses and with overhead power lines nearby, but this tree should only grow to 18 feet in both height and width so should fit nicely as a mature tree.  Info

 

Prairie Expedition Elm

A big tree for the big back yard.  Our neighbor Joyce has a nice big elm in her back yard and hopefully this one can fill in the skyline once that one has moved on.  This grows at a ridiculous rate, with many new 8' shoots each year.  They are too long and susceptible to wind damage so sometimes I cut the ends off.  This one was kind of a disorganized wreck when we got it, with a dead and sideways leader, so who knows if it will turn out to be a decent tree.  Maybe some good pruning tactics will become obvious as it continues to grow over the next few years.  Dutch-elm-disease resistant.  $280.  Info

 

Shrubs/flowers

Tiny Tortuga Turtleheads on each side of the front gate.  Mammoth Yellow Quill Daisies to the south of the gate.  Fulda Glow Sedum by the turtleheads.

Windows SSH client removes hmac-sha1 (and host key algorithm ssh-rsa) from defaults

I ran into trouble connecting to some old network gear this week.  It seems that the hmac-sha1 MAC was removed from the default client connection settings.  And then next I was having some trouble with the host key algorithm--the host key algorithm ssh-rsa was removed too.  They are still supported so can be specified manually in the client config file, such as this set that I use for older Extreme switches:

Host switch-1.domain.tld
    KexAlgorithms diffie-hellman-group1-sha1
    HostKeyAlgorithms ssh-dss
    Ciphers aes256-cbc
    MACs hmac-sha1
    HostKeyAlgorithms ssh-rsa

The errors I was getting:

Unable to negotiate with 192.168.1.1 port 22: no matching MAC found. Their offer: hmac-sha1,hmac-md5,hmac-sha1-96,hmac-md5-96

Unable to negotiate with 192.168.1.1 port 22: no matching host key type found. Their offer: ssh-rsa

You can see the difference in old and new ssh -vv outputs:

A Windows Server 2022 example:

OpenSSH_for_Windows_8.1p1, LibreSSL 3.0.2
<snip>
debug2: KEX algorithms: curve25519-sha256,curve25519-sha256@libssh.org,ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521,diffie-hellman-group-exchange-sha256,diffie-hellman-group16-sha512,diffie-hellman-group18-sha512,diffie-hellman-group14-sha256,diffie-hellman-group14-sha1,ext-info-c
debug2: host key algorithms: ecdsa-sha2-nistp256-cert-v01@openssh.com,ecdsa-sha2-nistp384-cert-v01@openssh.com,ecdsa-sha2-nistp521-cert-v01@openssh.com,ssh-ed25519-cert-v01@openssh.com,rsa-sha2-512-cert-v01@openssh.com,rsa-sha2-256-cert-v01@openssh.com,ssh-rsa-cert-v01@openssh.com,ecdsa-sha2-nistp256,ecdsa-sha2-nistp384,ecdsa-sha2-nistp521,ssh-ed25519,rsa-sha2-512,rsa-sha2-256,ssh-rsa
debug2: ciphers ctos: chacha20-poly1305@openssh.com,aes128-ctr,aes192-ctr,aes256-ctr,aes128-gcm@openssh.com,aes256-gcm@openssh.com
debug2: ciphers stoc: chacha20-poly1305@openssh.com,aes128-ctr,aes192-ctr,aes256-ctr,aes128-gcm@openssh.com,aes256-gcm@openssh.com
debug2: MACs ctos: umac-64-etm@openssh.com,umac-128-etm@openssh.com,hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,hmac-sha1-etm@openssh.com,umac-64@openssh.com,umac-128@openssh.com,hmac-sha2-256,hmac-sha2-512,hmac-sha1
debug2: MACs stoc: umac-64-etm@openssh.com,umac-128-etm@openssh.com,hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,hmac-sha1-etm@openssh.com,umac-64@openssh.com,umac-128@openssh.com,hmac-sha2-256,hmac-sha2-512,hmac-sha1

Freshly updated (2024-10-15) Windows 11 example:

OpenSSH_for_Windows_9.5p1, LibreSSL 3.8.2
<snip>
debug2: KEX algorithms: curve25519-sha256,curve25519-sha256@libssh.org,ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521,diffie-hellman-group-exchange-sha256,diffie-hellman-group16-sha512,diffie-hellman-group18-sha512,diffie-hellman-group14-sha256,ext-info-c,kex-strict-c-v00@openssh.com
debug2: host key algorithms: ssh-ed25519-cert-v01@openssh.com,ecdsa-sha2-nistp256-cert-v01@openssh.com,ecdsa-sha2-nistp384-cert-v01@openssh.com,ecdsa-sha2-nistp521-cert-v01@openssh.com,sk-ssh-ed25519-cert-v01@openssh.com,sk-ecdsa-sha2-nistp256-cert-v01@openssh.com,rsa-sha2-512-cert-v01@openssh.com,rsa-sha2-256-cert-v01@openssh.com,ssh-ed25519,ecdsa-sha2-nistp256,ecdsa-sha2-nistp384,ecdsa-sha2-nistp521,sk-ssh-ed25519@openssh.com,sk-ecdsa-sha2-nistp256@openssh.com,rsa-sha2-512,rsa-sha2-256
debug2: ciphers ctos: chacha20-poly1305@openssh.com,aes128-ctr,aes192-ctr,aes256-ctr,aes128-gcm@openssh.com,aes256-gcm@openssh.com
debug2: ciphers stoc: chacha20-poly1305@openssh.com,aes128-ctr,aes192-ctr,aes256-ctr,aes128-gcm@openssh.com,aes256-gcm@openssh.com
debug2: MACs ctos: umac-64-etm@openssh.com,umac-128-etm@openssh.com,hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,umac-64@openssh.com,umac-128@openssh.com,hmac-sha2-256,hmac-sha2-512
debug2: MACs stoc: umac-64-etm@openssh.com,umac-128-etm@openssh.com,hmac-sha2-256-etm@openssh.com,hmac-sha2-512-etm@openssh.com,umac-64@openssh.com,umac-128@openssh.com,hmac-sha2-256,hmac-sha2-512

The changes can be found in the myproposal.h file.  Below is a before and after of the MAC section since the diffs on GitHub seemed confusing to me.

Before:

#define    KEX_SERVER_MAC \
    "umac-64-etm@openssh.com," \
    "umac-128-etm@openssh.com," \
    "hmac-sha2-256-etm@openssh.com," \
    "hmac-sha2-512-etm@openssh.com," \
    "hmac-sha1-etm@openssh.com," \
    "umac-64@openssh.com," \
    "umac-128@openssh.com," \
    "hmac-sha2-256," \
    "hmac-sha2-512," \
    "hmac-sha1"

After:

#ifdef WINDOWS
#define    KEX_SERVER_MAC \
    "umac-64-etm@openssh.com," \
    "umac-128-etm@openssh.com," \
    "hmac-sha2-256-etm@openssh.com," \
    "hmac-sha2-512-etm@openssh.com," \
    "umac-64@openssh.com," \
    "umac-128@openssh.com," \
    "hmac-sha2-256," \
    "hmac-sha2-512,"
#else
#define    KEX_SERVER_MAC \
    "umac-64-etm@openssh.com," \
    "umac-128-etm@openssh.com," \
    "hmac-sha2-256-etm@openssh.com," \
    "hmac-sha2-512-etm@openssh.com," \
    "hmac-sha1-etm@openssh.com," \
    "umac-64@openssh.com," \
    "umac-128@openssh.com," \
    "hmac-sha2-256," \
    "hmac-sha2-512," \
    "hmac-sha1"
#endif

They did an #IFDEF so that this only affects the Windows client.

And the changes for the ssh-rsa setting were made way back in 2021 yet took until now to reach my PC.

Qotom 1U server SATA power+data cable

I have added a new node to my Proxmox cluster--a Qotom 1U rackmount device.  This came to my attention via a pretty thorough STH review.  This hits a sweet spot for me.  I want to have ~5 nodes in my cluster for Ceph reasons but don't have a ton of continuous load going on.  Because capacity and power draw will both be multiplied by five, I don't need a lot of compute/storage per node and want to keep power draw low.

Interlude on my disk selections

Ideally each would have one OS drive and two or three Ceph OSDs.  I am trying to do SSDs with power loss protection so selection is a bit limited.  But usually I can find used enterprise SSDs with this feature for approximately the same price as decent new consumer SSDs.

The system features two NVMe slots and two SATA ports.  I would have preferred OS on SATA and then fill both NVMe slots with OSD disks.  But it is difficult to find 80mm enterprise NVMe SSDs of 2TB or 4TB capacity. The Samsung PM983 would be perfect but it is a 22110 size (110mm long) and while I am always down for funky NVMe hold-downs, this case has a hard stop on length shortly after 80mm where the chassis wall is.

So I ended up with a Samsung MZ-7LH3T80 3.84TB disk which is SATA but will be fine.  And I picked up a Micron 7300 Pro 480GB as the OS disk (no need for an enterprise disk here but I didn't have any other 80mm units around so figured if I am going to be spending $45 anyways for a decent one I might as well do it).  Both used.

On with the main event

The server comes with one SATA power+data cable, and the power cable is non-standard.  The STH review used the cable from their second unit to hook up two SATA disks but no one sent me a second server!  Some searching revealed that this connector is described as a "PH2.0 small 4-pin" cable.  Or more specifically it is JST's PH series connector which features a 2.0mm pitch.

I was able to find a SATA assembly including the power and data cables on AliExpress.  But don't get that one.  The right-angle data plug is bent the "wrong" way and won't fit on the motherboard!  This one looks to be correct based on the picture (warning: the power pins aren't correct).

Not wanting to wait for a new delivery, I simply cut the disk-end connector in half with a bandsaw so that I could use only the power portion, and paired that with an on-hand old SATA data cable.

The computer would not boot with this hack-job connected.  So my first suspicion was the powercable pin order as the colors were different from those on the in-box cable.  I was able to rearrange the pins and things worked.

The drive side of the power cable has wires in this order: black, red, black, yellow.  I moved those first two (black and red) on the motherboard end to be similar to the cable that came with the server and it is running fine. (aka use the black wire that seems to be paired with the red)

The pins have a barb that holds them in the plastic connector.  You can poke or bend the barb with something tiny--I use the "SIM" bit in my iFixIt screwdriver kit.  And then try to bend the barb back out before re-seating the pin in the new home.

Proxmox/Ceph storage performance notes

I recently found the XtremeOwnage.com blog and love the content. I will write some similar stuff.

My homelab has been a Proxmox cluster with Ceph for almost a year now. It takes a lot of hardware and is complex but I love the flexibility to rearrange the storage. Between 5 nodes I have 10 SSDs as OSDs. Some are consumer-class which I now understand are quite harmful to performance. Most of the ~1TB ones are workplace "e-waste redirect" units so are Intel datacenter-class models that presumably have PLP capacitors.  I started out with 1gbps per node so performance was never going to be great. But now I have some nodes running 5x1gbps so I am getting more interested in performance.

Here are the OSDs:


 The test pool that I created for benchmarking:


Write performance:

root@proxmox1:~# rados bench -p testpool 300 write --no-cleanup
hints = 1
Maintaining 16 concurrent writes of 4194304 bytes to objects of size 4194304 for up to 300 seconds or 0 objects
Object prefix: benchmark_data_proxmox1_200202
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
    0       0         0         0         0         0           -           0
    1      16        53        37   147.986       148    0.583851    0.320801
    2      16        90        74   147.987       148    0.147199    0.325505
    3      16       135       119   158.654       180    0.138938    0.328477
<snip>
  294      16      8702      8686   118.165       152    0.293002    0.540915
  295      16      8716      8700   117.954        56    0.203445    0.540626
  296      16      8738      8722   117.853        88    0.501285    0.541629
  297      16      8759      8743   117.739        84    0.643936    0.542475
  298      16      8784      8768   117.679       100    0.112998    0.542322
  299      16      8804      8788   117.553        80    0.232327    0.542073
2024-09-19T22:20:13.695935-0500 min lat: 0.0476777 max lat: 4.72464 avg lat: 0.543161
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
  300      15      8824      8809   117.441        84    0.234312    0.543161
Total time run:         300.739
Total writes made:      8824
Write size:             4194304
Object size:            4194304
Bandwidth (MB/sec):     117.364
Stddev Bandwidth:       42.9952
Max bandwidth (MB/sec): 220
Min bandwidth (MB/sec): 4
Average IOPS:           29
Stddev IOPS:            10.7488
Max IOPS:               55
Min IOPS:               1
Average Latency(s):     0.545222
Stddev Latency(s):      0.463362
Max latency(s):         4.72464
Min latency(s):         0.0476777

Sequential read performance

root@proxmox1:~# rados bench -p testpool 100 seq
hints = 1
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
    0       0         0         0         0         0           -           0
    1      16        93        77   307.953       308   0.0125451    0.149726
    2      16       153       137   273.969       240   0.0473266    0.194905
    3      16       209       193   257.307       224   0.0672088     0.22432
    4      16       257       241   240.977       192   0.0402067    0.246331
    5      16       324       308   246.377       268    0.724575    0.241547
<snip>
   95      16      5348      5332   224.483       284    0.277306    0.283629
   96      16      5396      5380   224.145       192   0.0401958    0.283725
   97      16      5444      5428   223.813       192   0.0127102    0.284014
   98      16      5490      5474   223.407       184   0.0136518    0.284502
   99      16      5547      5531   223.453       228    0.151732    0.284663
2024-09-19T22:28:46.137589-0500 min lat: 0.00903257 max lat: 1.49518 avg lat: 0.2848
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
  100      15      5602      5587   223.458       224    0.141309      0.2848
Total time run:       100.495
Total reads made:     5602
Read size:            4194304
Object size:          4194304
Bandwidth (MB/sec):   222.976
Average IOPS:         55
Stddev IOPS:          7.95398
Max IOPS:             78
Min IOPS:             45
Average Latency(s):   0.28573
Max latency(s):       1.49518
Min latency(s):       0.00903257

Random read performance

root@proxmox1:~# rados bench -p testpool 100 rand
hints = 1
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
    0       0         0         0         0         0           -           0
    1      16       110        94   375.941       376    0.182853    0.128227
    2      16       214       198   395.945       416   0.0210152    0.148397
    3      16       303       287   382.622       356    0.492813     0.15927
    4      16       396       380   379.959       372   0.0994178    0.160321
<snip>
   95      16      8401      8385   353.019       372   0.0999113    0.180418
   96      16      8504      8488   353.633       412    0.177019    0.180185
   97      16      8604      8588   354.111       400   0.0284281    0.179882
   98      16      8703      8687   354.535       396    0.326019    0.179665
   99      16      8801      8785   354.913       392    0.341465    0.179431
2024-09-19T22:31:27.955888-0500 min lat: 0.00162297 max lat: 1.04212 avg lat: 0.179455
  sec Cur ops   started  finished  avg MB/s  cur MB/s last lat(s)  avg lat(s)
  100      16      8880      8864   354.524       316   0.0563875    0.179455
Total time run:       100.436
Total reads made:     8880
Read size:            4194304
Object size:          4194304
Bandwidth (MB/sec):   353.658
Average IOPS:         88
Stddev IOPS:          9.03027
Max IOPS:             112
Min IOPS:             63
Average Latency(s):   0.179924
Max latency(s):       1.04212
Min latency(s):       0.00162297

Three-way light sockets

I often have one or more lamps with three-way bulb sockets, but rarely bother purchasing a three-way bulb.  One might think that this results in some unpredictable behavior when trying to switch the light, as the next pull or turn may or may not turn it on or off.  But if you always turn two notches then the bulb will turn on / off on one of those notches and the behavior is predictable.

LeapFrog Fridge Phonics decoded

LeapFrog Fridge Phonics pieces front and back

The kids' toy "LeapFrog Fridge Phonics" has letter pieces that mount in a decoder, which then say different things depending on which letter you mounted.  Of course this requires some system for describing each letter, which I have documented below.

There are six bits.  I will mark the presence of a bump as 1, and no bump as a 0.  A couple pieces were missing but I am reasonably certain I have deduced those values correctly (guesses in parenthesis).  I have recorded them in what seems like "most-significant bit on the left" order.  If you have the letter upright and look at the back at the bumps along the bottom, this is the matching order.

All sequences ending in 00 were skipped.  And I suppose any starting with 00 were also skipped, since A starts off the sequence with the lowest binary value that would satisfy both criteria.  I'm thinking that the mechanics of the toy work best when the force is distributed on both ends of the pieces, hence always wanting at least one bump at or next to each end.

I find the two-code gap between N and O to be a real puzzle.  It must be for Ñ but I found a video of a Spanish version of the toy and the Ñ was just an aside drawn on the N.

The characters do actually line up with the characters in the ASCII binary table if:

  • Shift a digit in from the left.
  • Perform some modulo 3 adjustment to compensate for the gaps.
  • Compensate for the Ñ gap (module 14 adjustment?).
A(010001)
B 010010
C 010011
  010100
D(010101)
E 010110
F 010111
  011000
G 011001
H 011010
I 011011
  011100
J 011101
K 011110
L 011111
  100000
M 100001
N 100010
  100011
  100100
O 100101
P 100110
Q 100111
  101000
R 101001
S 101010
T 101011
  101100
U 101101
V 101110
W 101111
  100000
X 110001
Y 110010
Z 110011
  110100

Lint trap logic

Back when I had roommates I sometimes imagined that they would scorn me for not emptying the dryer lint trap after using the dryer. The logic behind my actions:

  • Messing with the lint trap could get lint or debris on the clean clothes (whether they are still in the dryer or nearby in a basket).
  • When you start a load, you really should check/verify the lint trap anyways. Might as well just clean it then.

In the Beginning was the Command Line by Neal Stephenson

Date completed
9 months ago

An interesting read, weaving between topics of operating system histories and philosophies.  I am about a year into my Linux journey and I think this mostly motivational for me to continue that, but also made me think what would have been different if had succeeded in my experiments to get Linux running at home in the basement during my high school years.  I remember burned Mandrake Linux CDs and also Ubuntu ones that I got in the mail for free.

Completion status
Rating

Migrating VMs from KVM/QEMU to Proxmox including Ceph

Here is my process.  Assume the VM is named zabbix, one of the Proxmox hosts is named proxmox1.example.com, the ID of your new VM is 1015, and your Ceph pool is named FirstPool.

  • Shut down the VM. (optionally change the interface name in /etc/network/interfaces to ens18 first)
  • Disable autostart on the VM: virsh autostart --disable --domain zabbix
  • Copy the qcow2 file to the Proxmox server.  A couple options, all run from the old KVM server (check source/destination order of scp command!):
    • To the Proxmox local disk: scp /data/vms/zabbix.qcow2 root@proxmox1.example.com:/var/lib/vz/images/zabbix.qcow2
    • Or maybe your VMs are big and your Proxmox disks are small.  You can copy directly to a CephFS if you have created one: scp /data/vms/zabbix.qcow2 root@proxmox1.example.com:/mnt/pve/cephfs/zabbix.qcow2
  • Import the VM into Proxmox: qm create 1015 --scsi0 FirstPool:0,import-from=/var/lib/vz/images/zabbix.qcow2 --boot order=scsi0  FirstPool is the Ceph pool.  It creates a disk in there automatically named vm-1015-disk-0.
  • Find the VM in the Proxmox web GUI and make some changes.  Probably many of these could be included in the previous step.
    • Add a network adapter.
    • Bump up CPU cores and RAM to whatever you want.
    • Change "Start at boot" to Yes.
    • Change the name of the VM (to "zabbix" for this example).
    • Change the OS type.
  • Boot the VM
  • If you didn't before shutting down, change the interface name in /etc/network/interfaces to ens18.  Make sure you check the interface name with ip a first to verify that it is ens18.  I have no idea why mine are always set to that but it seems consistent.

BTW copying to CephFS could be slower.  I have three nodes, all SSDs, and 1gbps connections and it seems to run at about 35MB/s versus 120MB/s that I was seeing to local disk.  Of course the biggest VMs need to be copied to slower storage!  But then that pain was dulled when I observed that my import speeds were over 1GB/s which is way faster than when importing from local storage.

Even though I think I have thin provisioning enabled in KVM, it seems to copy the full disk size across the network, and into the .qcow2 file..  But then when it imports into Ceph it seems to be thin provisioned again.  Smart but mysterious.

Seeing surprisingly high Ceph space usage afterwards?  You could be like me and assume that thin-provisioning is broken.  Or just go delete that .qcow2 (now that you have imported it) that you stuffed onto the CephFS which gets replicated 3x with no thin provisioning.