Why can't xenstored be restarted?

I'm looking for some definite info about when it's allowed to restart The various xen daemons (xend, xenstored and xendeskled ).
Up to now I worked with the assumption that I can restart them while Guest domains are running on the host and there's no harm done. Indeed there wasn't, until now, when on restart one of my Domains Became domain-unnamed (but kept on working), while the others Disappeared and didn't ping anymore...

I'm looking for some definite info about when it's allowed to restart the various xen daemons (xend, xenstored and xenreceivled ).

You cannot restart xenstored at all, it will break basic functionality. The other two can often be restarted but will cause issues sometimes.

You can't restart any qemu-DM processes.

Hi,

I'm looking for some definite info about when it's allowed to restart the various xen daemons (xend, xenstored and xenreceivled ).

Up to now I worked with the assumption that I can restart them while guest domains are running on the host and there's no harm done. indeed there wasn't, until now, when on restart one of my domains became domain-unnamed (but kept on working), while the others disappeared and didn't ping anymore...

(I 've got the domain info dump from XM dmesg and the xend logs if anybody can debug this. I'm using xen 3.2.1 .)

One thing you never (ever) want to re-start is xenstored. xend however can usually be re-started without issue.

Think of xenstored as something like "/proc as a service ". xen itself does not care about the name of your guest, it only needs its ID, flags, etc. if you look in/usr/include/xen/domctl. h, in particle the type xc_domaininfo_t, you'll see how much the userland tools rely on xenstore .. not just for storage, but watches. this keeps the bloat out of xen itself and in userspace where it belongs. that's one of the reasons why xen is a well-designed modern highly efficient microkernel.

To make the store persistent, please small writes/reads to dom-0's disk wocould be needed and wocould be a disaster. that's why its all done in memory. nevermind I/O wait and latency as far as I know, there is no way to export/import the contents of the store on a graceful re-start. trying to do the same while handling a segmentation fault wocould be at (BEST) questionable.

In all reality, I have seen xend crash, but never xenstored. the only practical reason for including stuff to dump and re-import its contents wocould be a Graceful Restart .. but why wocould you re-start it if its running fine?

 

Sorry for top posting, this comment in xenstore/xenstored_core.c is relevant:

/* XXX When we make xenstored able to restart, this will have to become cleverer, checking for existing domains and not removing the corresponding entries, but for now xenstored cannot be restarted without losing all the registered watches, which breaks all the backend drivers anyway. we can therefore get away with just clearing/local and expecting xend to put the appropriate entries back in.

When this change is made it is important to note that dom0's entries must be cleaned up on reboot _ before _ this daemon starts, otherwise the backend drivers and dom0's balloon driver will pick up stale entries. in the case of the balloon driver, this can be fatal. */

So, at least (For A While), the store is transient. I dug that up after double checking what -- preserve-local did. in all reality, you'll lose xend prior to losing xenstored.

 

Http://markmail.org/message/avqamaffh5zeqxav? Q = xenstored + restart # query: xenstored % 20 restart + Page: 1 + mid: zzw3ds7erzu7grqd + state: Results