On Tue, Jul 22, 2003 at 09:15:39AM -0400, Jason Edgecombe wrote:
Hi everyone,
I am having estimate timeouts from an amanda client.
The client's logs for amandad have the following error messages at the end: amandad: time 81.782: dgram_recv: timeout after 10 seconds amandad: time 81.782: waiting for ack: timeout, retrying amandad: time 91.782: dgram_recv: timeout after 10 seconds amandad: time 91.782: waiting for ack: timeout, retrying amandad: time 101.782: dgram_recv: timeout after 10 seconds amandad: time 101.782: waiting for ack: timeout, retrying amandad: time 111.782: dgram_recv: timeout after 10 seconds amandad: time 111.782: waiting for ack: timeout, retrying amandad: time 121.782: dgram_recv: timeout after 10 seconds amandad: time 121.782: waiting for ack: timeout, giving up! amandad: time 121.782: pid 17514 finish time Tue Jul 22 00:17:03 2003
My etimeout value on the server is -1800.
Doesn't seem like estimate timeouts, starting after only 81 seconds. Sounds like some network communication attempt failing.
I'm running amanda 2.4.3 on the client and 2.4.4 on the server.
My firewall is properly letting packets through.
Both client and server share a connection via the campus 100mbs network and a private gigabit network. I'm backing up over the gigabit network.
I can run the backup manually and the estimate phase runs fine, but it doesn't want to run at night.
Does anyone have any insight into this?
Two networks. Sounds like the amanda connection works over one but not the other. And for some reason, during the day, with manual attempts the working network is used, in evening, from cron, the other, non-working network is used. Maybe name resolution differences on the two networks?
I checked this with ethereal, everthing is in fact running over the gigabit ethernet.
I noticed that the amanda server uses udp port 993 as the src. so the first part of the amanda network conversation goes like this:
server port 993 -> client port 10080 hi server port 993 <- client port 10080 hi back server port 993 -> client port 10080 sendsize server port 993 -> client port 10080 ack sendsize
Is this correct? does amanda always use port 993 as the source port?
I hypothesize that my dumps are only finishing sometimes because port 993 is blocked by default. port 993 works work only a few minutes until the "allow related" rule timesout.
Jason Edgecombe
