Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Dimitrij Hilt wrote: >> Right. And now can you do >> >> dd if=<dev> skip=<start> bs=512 count=<size> iflag=direct | openssl md5 >> >> ... where <dev> is your backing device, <start> is the first out-of-sync >> sector as reported in the kernel log, and <size> is the number of >> out-of-sync sectors? >> > I did it allready. >> Please do that on both nodes, for a handful (say 5 or so) of out-of-sync >> areas reported in your syslog. >> >> If those MD5 sums match, then these are apparently false positives and >> we'll have to look into what's causing them. >> >> If, however, they do not match, replace "openssl md5" with "xxd" in the >> command above, and try to interpret those hex dumps. Are they completely >> different, do they not match at all, or are you seeing just one or two >> seemingly random differences? >> > Yes, i saw differences. These sectors was not false positives. OK. Do you have some of those xxd dumps still lying around? Can you post one or two pairs of those? > At the moment the problem is away. I did: > change meta-disc to internal > disable data-integrity-alg Did you try these changes one at a time, or both at once? Florian -- : Florian G. Haas : LINBIT Information Technologies GmbH : Vivenotgasse 48, A-1120 Vienna, Austria