iscsi target and VMware/esxi timeouts -- SOLVED

[Daniel Braniss]

> On 14 Nov 2017, at 11:28, Patrick M. Hausen <hausen at punkt.de> wrote:
> 
> Hello,
> 
>> Am 14.11.2017 um 10:08 schrieb Daniel Braniss <danny at cs.huji.ac.il>:
>> 
>> Hi,
>> we are experimenting issues with several esxi’s servers that use freebsd 10.2 stable as a iscsi target.
>> ie:
>> Nov 11 17:58:16 store-07 kernel: WARNING: 132.65.11.201 (iqn.1998-01.com.vmware:pe-02-2fa7cd9e): no ping reply (NOP-Out) after 5 seconds; dropping connection
>> Nov 11 17:58:16 store-07 kernel: WARNING: 132.65.11.201 (iqn.1998-01.com.vmware:pe-02-2fa7cd9e): no ping reply (NOP-Out) after 5 seconds; dropping connection
>> Nov 11 17:58:16 store-07 kernel: WARNING: 132.65.11.205 (iqn.1998-01.com.vmware:pe-03-13e8b52d): no ping reply (NOP-Out) after 5 seconds; dropping connection
>> Nov 11 17:58:17 store-07 kernel: WARNING: 132.65.11.203 (iqn.1998-01.com.vmware:pe-13-60e87d06): no ping reply (NOP-Out) after 5 seconds; dropping connection
>> Nov 11 17:58:17 store-07 kernel: WARNING: 132.65.11.205 (iqn.1998-01.com.vmware:pe-03-13e8b52d): no ping reply (NOP-Out) after 5 seconds; dropping connection
>> 
>> these are 3 different esxis that almost at the same time the target looses connection to the initiators.
>> at the moment most ‘clients’  recover from the scsi error, but older freebsds don’t.
>> 
>> in any case, increasing the timeout is not helping.
>> 
>> any clues are welcome :-)
>> 
>> over the weekend i’m planning to upgrade the target to 11.1 and take for the hills.
> 
> Are you using istgt or ctld?
> 
> We have did experience similar occasional problems with the former
> but never with the latter.
> 
> Patrick

the iscsi initiator of the esxi’s (VMware) does answer to NOP’s once in a blue moon!
this was checked by sniffing the network.

setting kern.cam.ctl.iscsi.ping_timeout=0 solved this.

thanks,
	danny