> > Solaris has live patching, but not live activation of kernel patches - > a reboot is required. > Most of the Illumos distros do the same.
Are there any document about these stuff? > Sun did live activation of kernel patches over 10 years ago (Project > DUKS - https://blogs.oracle.com/jimmo/entry/the_fate_of_hot_patching) > but never productized it. The hard part is not the kernel live patching, > but the management infrastructure that records precisely what you are > running, and that was beyond the capabilities of the packaging system > in use back then. Oracle has said they are revisiting this in Solaris, > now that Solaris has a packaging system which can handle it (IPS) and > Oracle have the Ksplice technology in their Linux. > In open source Linux world there are also the similar technologies like kpatch/kGraft. In fact, dynamic software updating (DSU) is just in research stage and not widely used in industry. "Retrofitting" current OS may not deliver thorough live patching which needs innovations in bundle of hardware(like mainframe) and kernel and programming languages. But any effort to minimize service outage is valuable. > > I think it is also the guideline for "zpool|zfs upgrade" command > > operation even I played them on-the-fly without any issues. But ZFS > is > > already a proven *enterprise* storage solution, we should definitely > > expect the minimum service outage just like NetApp/EMC. > > When you want that level of availability, clustering is normally used, > and ZFS works fine in clusters (at least in Solaris/Illumos - I don't > know about ZoL). ZFS is not native cluster file system. Cluster needs extra software stacks. HA is always a really touchy and tricky topic. It is application-specific and involves different levels of HA. What solution works well with ZFS in Soaris/Illumos? Sorry if it is off-topic and we may start a new thread. > You do not normally issue "zpool|zfs upgrade" commands as part of a > software upgrade, as it would prevent you from rolling back. It's > normally only done after period of stable running (and for a cluster, > after all nodes have been software upgraded), when you are sure you no > will longer need to roll back. Agree! I just want to use this to analogize/simulate the effect from patching. Thanks. Fred ------------------------------------------- illumos-discuss Archives: https://www.listbox.com/member/archive/182180/=now RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be Modify Your Subscription: https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4 Powered by Listbox: http://www.listbox.com
