To reply to my own message. I just noticed there was a recent iso, so we quick tested that to discount any oddball issue with pkg update –v’ing a 4/16 host to current. I should point out that a fresh ISO exhibits this behavior as well; again at version illumos-2016291. My concerns would be some sort of filesystem level metadata corruption that has been introduced somewhere between 8/16 and this point. I think I would be very concerned if I were creating filesystems in a mixed-version envirionment and expecting to replicate them.
I can provide more detail, but, basically these are vanilla fs’s created with defaults. We aren’t setting any properties; and I’ve even tried unsetting properties, such as compression and checksumming. The sends are vanilla zfs send, noptions to receives with only the –v option and a target. Does anyone have any input/ideas? Thanks, Joe On 11/1/16, 12:57 PM, "Hetrick, Joseph P" <[email protected]> wrote: Hey folks, We’ve recently discovered a concerning issue. Our environment has several version of OI from 151a6 up to illumos-2816291. We discovered that zfs sends between the 2816291 systems and systems at least to 151a8 will fail with the recv end hanging. This causes some sort of race state, and eventually the system needs to be kicked. We have a couple illumos-a7317ce systems that do not exhibit this behavior when they are receivers. What we found with some experimentation is: Take bare 4/16 image and create pool/filesystem, send to OI systems; no problem. Update 4/16 image to current; create filesystem, send to OI system: rcvr hang/failure. Reboot into 4/16 pre-update BE, import current created pool create FS on current created pool, send dataset to OI; no problems. Filesystem and pool properties/versions show no change between them; At the moment our plan is to revert back to 4/16, but this leaves us unable to pull packages from the IPS repo. I have not extensively trolled the illumos bug tracker to find a match for this issue; but if it is known, I apologize. Has anyone experienced this on the list? Thanks, Joe _______________________________________________ openindiana-discuss mailing list [email protected] https://openindiana.org/mailman/listinfo/openindiana-discuss _______________________________________________ openindiana-discuss mailing list [email protected] https://openindiana.org/mailman/listinfo/openindiana-discuss
