When I leave large/fast transfers running for awhile (i.e. FTP sessions between 2 local GigE-equipped PCs) I see this in the logs fairly often, and the FTP transfers stop (but the FTP program eventually retries):
I know it's just a warning, and things like samba and FTP recover, but is it something I should be worried about? It seems to make the connection hiccup. I get it when I have the systems connected through a GigE switch, and also when I connect them together directly.
I don't know what a TCB is, but is it something I can tune the kernel to have more of?
Thanks!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
...when I actually started to graph my transfers (using MRTG) I can see that:
a) Speeds would slow over time (i.e. 10MB/s FTP would be at 100KB/s in an hour
b) The connection occasionally totally stops (the interface is dead: you cannot ping, no traffic). If I unplug the cable from the interface and plug it back in again it restarts at full speed. I've tried 2 different non-et131x interfaces on the second PC (Nvidia and Broadcom), as well as putting a switch between them and it's the et131x that always need the reset (i.e. with the switch in place, the switch->et131x cable needs to be reset... the switch->PC#2 cable makes no difference when reset.
This is driving me crazy: maybe I'm just unlucky.... but when I drive low/slow traffic through any of my 3 et131x cards they are fine... lots of traffic and they get flakey.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Even stranger... when the connection stops it's affecting all 3 interfaces (i.e. the transfer rate graphs for all 3 drop to zero). Why would unplugging/plugging-in the ethernet cable on eth1 also affect eth2 and eth3?
Maybe one of the 3 cards is bad: I'll try shuffling them around in different slots to see if the problem follows one card around.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Ooops... all 3 may be affected because when one interface goes down the routing tables get altered and flushed... so apps that were balanced over the 2 internet connections all have to reconnect etc. So maybe it is just a problem with one card.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
When I leave large/fast transfers running for awhile (i.e. FTP sessions between 2 local GigE-equipped PCs) I see this in the logs fairly often, and the FTP transfers stop (but the FTP program eventually retries):
et131x.ko:WARNING:et131x_send_packet Can't obtain a TCB
et131x.ko:WARNING:et131x_send_packets Resources problem, Queue tx packet
I know it's just a warning, and things like samba and FTP recover, but is it something I should be worried about? It seems to make the connection hiccup. I get it when I have the systems connected through a GigE switch, and also when I connect them together directly.
I don't know what a TCB is, but is it something I can tune the kernel to have more of?
Thanks!
Also, maybe related, maybe not...
...when I actually started to graph my transfers (using MRTG) I can see that:
a) Speeds would slow over time (i.e. 10MB/s FTP would be at 100KB/s in an hour
b) The connection occasionally totally stops (the interface is dead: you cannot ping, no traffic). If I unplug the cable from the interface and plug it back in again it restarts at full speed. I've tried 2 different non-et131x interfaces on the second PC (Nvidia and Broadcom), as well as putting a switch between them and it's the et131x that always need the reset (i.e. with the switch in place, the switch->et131x cable needs to be reset... the switch->PC#2 cable makes no difference when reset.
This is driving me crazy: maybe I'm just unlucky.... but when I drive low/slow traffic through any of my 3 et131x cards they are fine... lots of traffic and they get flakey.
Even stranger... when the connection stops it's affecting all 3 interfaces (i.e. the transfer rate graphs for all 3 drop to zero). Why would unplugging/plugging-in the ethernet cable on eth1 also affect eth2 and eth3?
Maybe one of the 3 cards is bad: I'll try shuffling them around in different slots to see if the problem follows one card around.
Ooops... all 3 may be affected because when one interface goes down the routing tables get altered and flushed... so apps that were balanced over the 2 internet connections all have to reconnect etc. So maybe it is just a problem with one card.