On April 18, 2020 12:04:41 AM UTC, Bob Friesenhahn <[email protected]> wrote: >On Sat, 18 Apr 2020, Gergő Doma wrote: > >> Hi. >> >> autom4te tries to use flock() over NFS but NFS doesn't supports that. > >If this is so, why do OpenIndiana oi_151a9, Solaris 10, and Ubuntu >18.04 not report a problem? > >Is this an Illumos-specific locking agent issue, or is it a NFS >general issue? > >I do recall that Illumos re-wrote the NFS locking daemon. > >Bob
This sounds reminiscent of an issue I had a decade or more ago and discovered the mess of flock vs fnctl vs lockf on NFS and different clients around it. And I remember Sun vs Sun server/clients working well while with Linux could YMMV. In the end I may have adapted some in-house scripts to prefer fnctl even when given a choice of APIs. (Nowadays I have an OI VM as main workstation and storage, with a Linux VM build system and Windows hypervisor using it over NFS quite well... with rare issues, most notably from Java mvn-tests seeing .nfsXXXX files after cleaning workspaces). https://perlmonks.org/index.pl?node_id=69883 https://unix.stackexchange.com/questions/1777/flock2-versus-fcntl2-over-a-nfs Many links I found stress the use of lockd to have the locking at all. So the rewrite might play a role too. And a mix of NFS v2/3/4 supports might vary a lot - see if downgrading to v3 (forcing vers in automounter settings) helps for example. Jim -- Typos courtesy of K-9 Mail on my Android ------------------------------------------ illumos: illumos-discuss Permalink: https://illumos.topicbox.com/groups/discuss/T62b11c36b038c208-Me85691f02270c2a18d33391b Delivery options: https://illumos.topicbox.com/groups/discuss/subscription
