Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Nov 22 01:51

    dependabot[bot] on pip

    (compare)

  • Nov 22 01:51
    dependabot[bot] closed #575
  • Nov 22 01:51
    dependabot[bot] commented #575
  • Nov 22 01:51
    dependabot[bot] labeled #586
  • Nov 22 01:51
    dependabot[bot] opened #586
  • Nov 22 01:51

    dependabot[bot] on pip

    Bump pillow from 4.2.0 to 9.3.0… (compare)

  • Aug 19 16:10
  • Jul 06 19:50

    dependabot[bot] on pip

    (compare)

  • Jul 06 19:50
    dependabot[bot] closed #566
  • Jul 06 19:50
    dependabot[bot] commented #566
  • Jul 06 19:50
    dependabot[bot] labeled #585
  • Jul 06 19:50
    dependabot[bot] opened #585
  • Jul 06 19:50

    dependabot[bot] on pip

    Bump lxml from 3.6.4 to 4.9.1 i… (compare)

  • Jul 01 11:52
    dependabot[bot] labeled #584
  • Jul 01 11:52
    dependabot[bot] opened #584
  • Jul 01 11:52

    dependabot[bot] on pip

    Bump openpyxl from 2.3.5 to 2.4… (compare)

  • May 26 12:04
    marcheschi commented #583
  • May 26 11:49
    marcheschi commented #583
  • May 26 11:44
    marcheschi edited #583
  • May 26 11:39
    marcheschi opened #583
klebed
@klebed:matrix.org
[m]
2x E5-2697v2
toybomb
@toybomb:matrix.org
[m]
i wonder if its something with the CPU or the NIC
NIC would make more sense
klebed
@klebed:matrix.org
[m]
If it's NIC it would affect all VMs, I believe. I had that hypothesis, but it's ulikely.
CPU - might be. But test results say that somehow KVM and zones handle that well.
toybomb
@toybomb:matrix.org
[m]
well the virtualized NIC is different no?
1 reply
I see
klebed
@klebed:matrix.org
[m]
Does someone knows how to debug guakamole accessibility? We have issue with accessing KVM console on couple of nodes. All VMs on that nodes showing "please wait..." indefinitely.
Jan Poctavek
@YanChii
every VM has kvm_port property defined... get it using vmadm ad connect to that port using VNC client
guacamole service has logs inside mgmt VM
2 replies
klebed
@klebed:matrix.org
[m]
guacd
Jan Poctavek
@YanChii
yes
klebed
@klebed:matrix.org
[m]
only these rows in this log
sep 28 14:12:28 mgmt01.local guacd[2831]: Guacamole connection closed during handshake
sep 28 14:12:28 mgmt01.local guacd[2831]: guacd[2831]: INFO: Guacamole connection closed during handshake
Jan Poctavek
@YanChii
you can increase loglevel in conf... in etc
and restart the service
klebed
@klebed:matrix.org
[m]
but thanks... I think I have some clues where to dig
Jan Poctavek
@YanChii
guacd is quite stable... I've seen only a few problems with it
klebed
@klebed:matrix.org
[m]
When trying to connect through VNC directly it fails... just says that VNC server closed connection. I wonder what it could B
is there any logs of VNC server on nodes?
Jan Poctavek
@YanChii
the VNC is integrated into qemu and bhyve... no dedicated server exists
vmadmd just does port forward
bhyve has very barebone VNC implementation, not compatible with all clients
try different VMs, even those thatbwork over guacamole and compare
maybe the problematic VMs just need a stop/start
klebed
@klebed:matrix.org
[m]

Hi, colleagues!
Another topic. You ship OPNsense appliance and encourage to use it instead of sun access zone, but it has major flaw in UFS, which causes machine to become broken literally any time, especially if you reboot it.

Please look at the following quotation:
"pfSense filesystem corruption after power outage

pfSense has a serious flaw that has affected some of our customers. The UFS filesystem used by pfSense is sensitive to sudden power loss. When pfSense system losses power at a wrong moment, the UFS filesystem may get corrupted, which may result in unbootable router. This problem is not related to hardware, it affects all routers with pfSense.

Most customers that reported this problem to us said that it appeared right aftter a sudden electricity outage. One customer reported that it happened during a pfSense upgrade.

We have tried to replicate this by pulling the plug while the router was running, 15 times in a row, and managed to reproduce it once.

There are many threads on pfSense forum that decriebe exactly the same problem: unbootable or boot-looping operating system after electricity loss.

pfSense team is aware of this issue and recently added ability to choose ZFS instead of UFS when installing the OS. In our experience ZFS is immune to the issue, so we strongly recommend using ZFS instead of UFS."

Same allies to Opnsense as well. So it's worth to ship it with ZFS.

klebed
@klebed:matrix.org
[m]

:point_up: Edit: Hi, colleagues!
Another topic. You ship OPNsense appliance and encourage to use it instead of sun access zone, but it has major flaw in UFS, which causes machine to become broken literally any time, especially if you reboot it.

Please look at the following quotation:
"pfSense filesystem corruption after power outage

pfSense has a serious flaw that has affected some of our customers. The UFS filesystem used by pfSense is sensitive to sudden power loss. When pfSense system losses power at a wrong moment, the UFS filesystem may get corrupted, which may result in unbootable router. This problem is not related to hardware, it affects all routers with pfSense.

Most customers that reported this problem to us said that it appeared right aftter a sudden electricity outage. One customer reported that it happened during a pfSense upgrade.

We have tried to replicate this by pulling the plug while the router was running, 15 times in a row, and managed to reproduce it once.

There are many threads on pfSense forum that decriebe exactly the same problem: unbootable or boot-looping operating system after electricity loss.

pfSense team is aware of this issue and recently added ability to choose ZFS instead of UFS when installing the OS. In our experience ZFS is immune to the issue, so we strongly recommend using ZFS instead of UFS."

Same applies to Opnsense as well. So it's worth to ship it with ZFS.

Jan Poctavek
@YanChii
thanks @klebed:matrix.org for pointing this out. Having ZFS is good idea. We might need some tuning done to be able to fit ZFS in 512M RAM. Meanwhile you can turn on frequent snapshots on tne OPNSense VM and if the disaster strikes, you still have very high probability that most of the snapshots are consistent enough.
6 replies
wilbury
@wilbury:matrix.org
[m]
opnsense can be installed on zfs, ofc. but as time goes by, the UFS problems have been solved (there were quite a few UFS related fixes in freebsd recentlty)
1 reply
klebed
@klebed:matrix.org
[m]
It takes time to restore machine with opnsense but it's not that important. Worst thing is that it dies unexpectedly and then people can't use it for some time.
Jan Poctavek
@YanChii
restore from backup yes... snapshot rollback should be immediate. But I agree that the most inconvenient part is the need of restoring.
Jan Poctavek
@YanChii
FYI: danube.cloud domain is unavailable... DNS registrar problem, working on it
Paolo Marcheschi
@marcheschi
Thank you @YanChii
wilbury
@wilbury:matrix.org
[m]
ohai!
Jan Poctavek
@YanChii
we're back up since yesterday
wilbury
@wilbury:matrix.org
[m]
@YanChii: great.
Paolo Marcheschi
@marcheschi
Hi what about the Idea to update the platform?
Jan Poctavek
@YanChii
yes, it's needed... @wilbury:matrix.org is taking over the platform so we'll meet in close future
Paolo Marcheschi
@marcheschi
perfect !!
wilbury
@wilbury:matrix.org
[m]
yeah.
usama-nrk
@usama-nrk
Not sure how active this project is. Good idea to deploy? I loved SmartOS before Samsung. Not sure how active everything is now/
Jan Poctavek
@YanChii
Hi @usama-nrk. Illumos project is actively developed and many companies are contributing to it's development.
smartos changes are a bit sparser lately... but it's far from dead
you need to decide on your own
Jan Poctavek
@YanChii
regarding DC itself, it's doing it's job. You can try it and see if the features suit your needs. If you change your mind later, you can still dd your VMs to different platform.
usama-nrk
@usama-nrk
fair enough. I will give it a try. Hopefully all will be well.
May I ask what do you suggest for tiered storage, nvme, ssd and rust
Jan Poctavek
@YanChii
not sure I get your question. Are you trying to combine multiple device types into one storage pool? Is it in connection with DC?
With ZFS, you can add SSDs as write cache (ZIL) and read cache (L2ARC) to speed up the pool consisting of rotational disks.
Or you can create multiple pools that have different speeds (fast pool, slow pool, slow pool accelerated by ZIL and L2ARC, etc)