Time sync problem on gaia Open Server in Hyper-V

[Martin_Vilim]

Hello,

I started with Checkpoint before 12 months, with Security GW (Open Server VmWare) and Smart. The Smart and management are installed on the same server. Gaia runs on Hyper-V Windows 2012 R2. The virtual server version is 1. On the revision R80.10 Take 42

I experienced huge problem with time sync. Because of incompatibility between Gaia and Hyper-V described sk105862

I think it's not necessary to describe how bad is when the logs are time shifted.

Can you share your solution, how are you dealing with this issue? I found the NTP has a problems too. The solution from support is not clear, it was try and check. I'm wonder for your solutions of this issue, I think it may be different.

Thank you, Martin

[Danny]

In a Hyper-V environment, the Virtual Machine's clock moves faster than the hardware time because the Hyper-V clocksource does not work on 64-bit kernel (sk105862).

Therefore Hyper-V integration services should not offer time sync. Uncheck this setting in the "Advanced" section of hardware properties of the Hyper-V VM. Check Point Support will provide you a custom hotfix for your environment, for which you need to open a Service Request. In the mean time you can have a cronjob correcting the time every minute as Check Point suggests or configure Advanced NTP features on Gaia OS in order to do a NTP time sync with an in-house NTP server of yours every 16 seconds. If you need a shorter poll interval, use ntpdate.

NTP Troubleshooting

[Martin_Vilim]

I know about this sk, I put it in my question. I remember that I did with technical support a lot more things, some play with NTP, editing grub.conf... So I wonder what's work for other peoples. BTW time sync issue on the supported configuration, on logging system is horrible.

[Nikhil_Deshmukh]

Hotfix worked like a charm

Go #CheckPoint Go #CheckMates 

[Hatime_CHIKHI]

Did you have to uncheck the sync option in the "Advanced" section of hardware properties of the Hyper-V VM ?

We have installed the hotfix, it worked for a while but then again it stopped synchronizing correctly.

[Hatime_CHIKHI]

Update:

Checkpoint Support provided an other hotfix package, and the problem was solved after installing this new package.

[Nikhil_Deshmukh]

True Hatime CHIKHI‌,

Previously provided hotfix didn't work. Sorry for earlier.

I again contacted Check Point for the same, and they suggested by TAC is to go with On-going Jumbo Hotfix Take 131 or to create Cron Job on Management Server.

We went ahead with Cron Job as going with the On-going Take is a bit

[Steve_Payne1]

any chance of sharing your con job details so i can apply to my setup

thanks

[Nikhil_Deshmukh]

Hi Steve Payne‌,

• uncheck the sync option in the "Advanced" section of hardware properties of the Hyper-V VM
• Here you go, I have received the same from TAC.

# echo "*/1 * * * * root /sbin/hwclock --utc --hctosys" > /etc/cron.d/synctime
# chmod 644 /etc/cron.d/synctime
# service crond reload

• Give it sometime and then it will start working

# clish -c "show time"

[Steve_Payne1]

thank you,  as it was causing issues with out endpoint VPN's i exported from hyperv  and imported to VMware, been perfect since

[Benoit_Verove · Accepted solution]

Hi,

We had a similar issue with a R80.10 management on Hyper-V a few month ago.

We raised a service request. First, the TAC proposed us a workaround with a similar script to add to cron table.

Then we add a remote session with the support engineer : After backing it up, we have updated the grub with a value to change how the clock is being handled by the OS.
From then, no more clock problems.

Here is the value :
"notsc divider=10 clocksource=acpi_pm"

Regards,

Benoit

[Martin_Vilim]

Hi,

from my experience, the NTP sync in cron never works well. It caused bigger time mess in the logs.

The result that works well is the grub config update.

/etc/grub.conf

On the line with kernel parameters, I have divider=10 nophet notsc. Similar like Benoit wrote.

Regards,

Martin

[GG27]

The same issue affected a customer enviroment where HyperV is the HyperVisor used for their system.

The clock moved continuosly back and forward.

After a long TAC investigation, it provided me the workaround where I scheduled the sync between the host motherboard clock and the Gaia operating system.

[Yamil_Karim]

Do you think this would apply to Azure VMs? We are experiencing the same sort of issue on an Azure VM of this.