Hi, Gene Heskett wrote: > lsblk, which I've published several times, shows 5 drives.
Duh. Obviously this thread overstretches my mental capacity. > And I've since tried cp in addition to rsync, does the same thing, killing > the sysytem with the OOM but much quicker. cp using all system memory (32Gb) > in 1 minute, another 500K into swap adds another 15 secs, and the OOM kills > the system. So both cp and rsync act broken. I get the suspicion that your disk set overstretches the mental capacity of the hardware or the operating system. Both "cp" and "rsync" are heavily tested by the GNU/Linux community and quite independently developed. A common memory leak would have to sit deeper in the software stack, i.e. in kernel or firmware. > rsync, with a --bwlimit=3m set, takes much longer to kill the system but the > amount of data moved is very similar, 13.5G from clean disk to system freeze > for rsync, 13.4G for cp. This observation might be significant. But i fail to make up a theory. > gene@coyote:~/src/klipper-docs$ lsblk -d -o NAME,MAJ:MIN,MODEL,SERIAL,WWN > /dev/sd[hijkl] > NAME MAJ:MIN MODEL SERIAL WWN > sdh 8:112 Gigastone SSD GSTD02TB230102 > sdi 8:128 Gigastone SSD GST02TBG221146 > sdj 8:144 Gigastone SSD GST02TBG221146 > sdk 8:160 Gigastone SSD GSTG02TB230206 > sdl 8:176 Gigastone SSD GSTG02TB230206 This is just weird. I still have difficulties to believe that any disk manufacturer would hand out disks with colliding serial numbers. I googled for this phenomenon, but except two mails of Gene nothing similar popped up. One of these mails from a thread in december reveals that the three unique serial numbers GSTD02TB230102, GST02TBG221146, GSTG02TB230206 each come with a different version of "1C0", "7A0", "5A0", respectively. https://www.mail-archive.com/debian-user@lists.debian.org/msg799307.html That's unexpected, too, as the disk properties look identical elsewise. I guess that it is not possible to identify which disk came with which of the two separate purchases ? How many days were these purchases apart ? David Christensen wrote: > I suggest removing one GST02TBG221146 and one GSTG02TB230206. Put them on > the shelf, in other computer(s), or sell them. Then perhaps copying the > /home RAID10 2 TB to one Gigastone 2 TB SSD would work. I join this proposal. ... and dimly remember to have seen the proposal to attach the disks one by one without the other four, in order to see whether the serial numbers are the same as with all five together. Since you got quite some hardware zoo: Consider to try the Gigastone disks with a different machine. Do the serial numbers show up as with the machine where you experience all those difficulties. Have a nice day :) Thomas
