
thomasm at example
May 18, 2000, 6:37 AM
Views: 777
Permalink
|
|
Turning off the master cause the slave to freeze ..
|
|
Previously on another channel : >One more time I disturb you for DRBD... >I done the following test several times and got this : >- power off primary (rude) >- on the fly change the secondary to become primary. > >Then the computer "halt" or more exactly hang, freeze >Do you know which drsetup command to tape to fix it ? Be this I meaned to avoid the freeze as it seems to be due to a timeout in the communication protocol. As I am using the sync option on my partition will the A protocol avoid that ? > Please, could you give us/me a few more details. > E.g. Kernel versions, DRBD versions. This is a redhat 6.2 with the 2.2.14-12 kernel, I am using drbd 0.5.5 > Was just the IO system blocked (and still response to mouse & keyboard), or was it completely frozen? It was a complete freeze ... I don't have the magic keys enable in the kernel so I can't tell you more > What protocol was used? Protocol B > Any messages in syslog? accept failed -512 ack timeout detected (normal the other side is down) accept failed -512 At reboot : set_blocksize : b_count 1, dev drbd(43,0), block 851968! >Was there a FS mounted on top of DRBD? ext2 > PS: There is a new mailinglist, you are welcome to send any comments regarding DRBD to that list. ;*) I noticed too that after I was enable to reconnect my two nodes .. I was able to have two slave 1 slave 1 master but in this case the slave was not connected (a WFconnection on the master and unconnected on the slave - from memory) I had to mkfs.ext2 the underlying partition on the slave, reboot both to be able to use mkfs on my /dev/nb0 again .. I am installing a production mail server tommorrow using drbd to replicate on a auxiliary server with some piece of script to takeover and swap service. Any information before I do the switch ? Thomas
|