3

We are running ZFS on linux with centos 7.2. The array has 26 6tb Seagate sas drives in a mirrored configuration and then stripped across the mirrors to form our pool. 128gb arc, 480GB l2arc PCI flash card. Slog is supplied by a sata ssd. We are then sharing this out using samba. We get great write performance, decent read though slow on 100,00k+ small file reads. We have enable async io on smaba to help with this. During sync writes the reads of this small file data slows to a crawl. Can't figure out why or if there is any settings on zfs or samaba that I need to manipulate to alleviate this. Please and thank you for any help you my provide.

ZFS get all http://pastebin.com/SL0Sma1h

user2997301
  • 31
  • 2
  • Definitely. Can you give us some detail on your pool? I'd like to know the ashift setting, the make/model of SSD, your ZFS filesystem settings and anything else that could be relevant. Maybe just post "zpool get all" and "zfs get all pool/filesystem" for the filesystem you're concerned about. – ewwhite Jun 21 '16 at 20:28
  • For EL7, you'll want "xattr=sa" enabled. – ewwhite Jun 21 '16 at 20:30
  • I have adde dour zfs get all for you to look at. We do have xattr set to sa – user2997301 Jun 23 '16 at 15:15

0 Answers0