| Message ID | 20190314211210.7454-1-smayhew@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | nfs4.1: add a bunch of reboot tests | expand |
On Thu, Mar 14, 2019 at 05:12:06PM -0400, Scott Mayhew wrote: > (Note: These patches go on top of "nfs4.1: clean up the session and > client created in Environment.init()" from Jeff Layton sent on March > 14th). That was yours, wasn't it? Applied, anyway. > These patches add the following reboot tests: Those look like good ideas, I've applied the patches. The one thing I notice on a quick skim is that REBT5 depends on Linux knfsd-specific behavior. That's worth noting in a comment. (Also maybe we should add a flag to such tests. "knfsd"? We've got some tests flagged "ganesha".) --b. > > REBT2x: Server reboot with N clients reclaiming opens. > REBT3x: Server reboot with N clients reclaming opens. Server reboots > again after half the clients have reclaimed. > REBT4x: Server reboot with N clients reclaiming opens. Half the > clients attempt to reclaim twice. > (where x in {a, b, c}, with 'a' using 10 clients, 'b' using 100 clients, > and 'c' testing 1000 clients) > REBT5: Server reboot where the client starts reclaiming shortly before > the end of grace. > > Note all of these tests require the use of a helper script or manual > intervention to restart the server. > > Scott Mayhew (4): > nfs4.1: add some reboot tests > nfs4.1: add some more reboot tests > nfs4.1: still more reboot tests > nfs4.1: test delayed reclaim following a server reboot > > nfs4.1/server41tests/st_reboot.py | 284 +++++++++++++++++++++++++++++- > 1 file changed, 278 insertions(+), 6 deletions(-) > > -- > 2.17.2
On Thu, 14 Mar 2019, J. Bruce Fields wrote: > On Thu, Mar 14, 2019 at 05:12:06PM -0400, Scott Mayhew wrote: > > (Note: These patches go on top of "nfs4.1: clean up the session and > > client created in Environment.init()" from Jeff Layton sent on March > > 14th). > > That was yours, wasn't it? Applied, anyway. No, my patch modified Environment.finish(). Jeff's modifies Environment.init(). You'll need his patch too or these tests won't work. -Scott > > > These patches add the following reboot tests: > > Those look like good ideas, I've applied the patches. > > The one thing I notice on a quick skim is that REBT5 depends on Linux > knfsd-specific behavior. That's worth noting in a comment. > > (Also maybe we should add a flag to such tests. "knfsd"? We've got > some tests flagged "ganesha".) > > --b. > > > > > REBT2x: Server reboot with N clients reclaiming opens. > > REBT3x: Server reboot with N clients reclaming opens. Server reboots > > again after half the clients have reclaimed. > > REBT4x: Server reboot with N clients reclaiming opens. Half the > > clients attempt to reclaim twice. > > (where x in {a, b, c}, with 'a' using 10 clients, 'b' using 100 clients, > > and 'c' testing 1000 clients) > > REBT5: Server reboot where the client starts reclaiming shortly before > > the end of grace. > > > > Note all of these tests require the use of a helper script or manual > > intervention to restart the server. > > > > Scott Mayhew (4): > > nfs4.1: add some reboot tests > > nfs4.1: add some more reboot tests > > nfs4.1: still more reboot tests > > nfs4.1: test delayed reclaim following a server reboot > > > > nfs4.1/server41tests/st_reboot.py | 284 +++++++++++++++++++++++++++++- > > 1 file changed, 278 insertions(+), 6 deletions(-) > > > > -- > > 2.17.2
On Thu, Mar 14, 2019 at 07:18:33PM -0400, Scott Mayhew wrote: > On Thu, 14 Mar 2019, J. Bruce Fields wrote: > > > On Thu, Mar 14, 2019 at 05:12:06PM -0400, Scott Mayhew wrote: > > > (Note: These patches go on top of "nfs4.1: clean up the session and > > > client created in Environment.init()" from Jeff Layton sent on March > > > 14th). > > > > That was yours, wasn't it? Applied, anyway. > > No, my patch modified Environment.finish(). Jeff's modifies > Environment.init(). You'll need his patch too or these tests won't > work. Whoops, sorry. I don't see that mail from Jeff anywhere. Would someone mind resending? --b. > > -Scott > > > > > > These patches add the following reboot tests: > > > > Those look like good ideas, I've applied the patches. > > > > The one thing I notice on a quick skim is that REBT5 depends on Linux > > knfsd-specific behavior. That's worth noting in a comment. > > > > (Also maybe we should add a flag to such tests. "knfsd"? We've got > > some tests flagged "ganesha".) > > > > --b. > > > > > > > > REBT2x: Server reboot with N clients reclaiming opens. > > > REBT3x: Server reboot with N clients reclaming opens. Server reboots > > > again after half the clients have reclaimed. > > > REBT4x: Server reboot with N clients reclaiming opens. Half the > > > clients attempt to reclaim twice. > > > (where x in {a, b, c}, with 'a' using 10 clients, 'b' using 100 clients, > > > and 'c' testing 1000 clients) > > > REBT5: Server reboot where the client starts reclaiming shortly before > > > the end of grace. > > > > > > Note all of these tests require the use of a helper script or manual > > > intervention to restart the server. > > > > > > Scott Mayhew (4): > > > nfs4.1: add some reboot tests > > > nfs4.1: add some more reboot tests > > > nfs4.1: still more reboot tests > > > nfs4.1: test delayed reclaim following a server reboot > > > > > > nfs4.1/server41tests/st_reboot.py | 284 +++++++++++++++++++++++++++++- > > > 1 file changed, 278 insertions(+), 6 deletions(-) > > > > > > -- > > > 2.17.2
On Thu, Mar 14, 2019 at 09:00:36PM -0400, J. Bruce Fields wrote: > On Thu, Mar 14, 2019 at 07:18:33PM -0400, Scott Mayhew wrote: > > On Thu, 14 Mar 2019, J. Bruce Fields wrote: > > > > > On Thu, Mar 14, 2019 at 05:12:06PM -0400, Scott Mayhew wrote: > > > > (Note: These patches go on top of "nfs4.1: clean up the session and > > > > client created in Environment.init()" from Jeff Layton sent on March > > > > 14th). > > > > > > That was yours, wasn't it? Applied, anyway. > > > > No, my patch modified Environment.finish(). Jeff's modifies > > Environment.init(). You'll need his patch too or these tests won't > > work. > > Whoops, sorry. > > I don't see that mail from Jeff anywhere. Would someone mind resending? Never mind, I found it.--b. > > --b. > > > > > -Scott > > > > > > > > > These patches add the following reboot tests: > > > > > > Those look like good ideas, I've applied the patches. > > > > > > The one thing I notice on a quick skim is that REBT5 depends on Linux > > > knfsd-specific behavior. That's worth noting in a comment. > > > > > > (Also maybe we should add a flag to such tests. "knfsd"? We've got > > > some tests flagged "ganesha".) > > > > > > --b. > > > > > > > > > > > REBT2x: Server reboot with N clients reclaiming opens. > > > > REBT3x: Server reboot with N clients reclaming opens. Server reboots > > > > again after half the clients have reclaimed. > > > > REBT4x: Server reboot with N clients reclaiming opens. Half the > > > > clients attempt to reclaim twice. > > > > (where x in {a, b, c}, with 'a' using 10 clients, 'b' using 100 clients, > > > > and 'c' testing 1000 clients) > > > > REBT5: Server reboot where the client starts reclaiming shortly before > > > > the end of grace. > > > > > > > > Note all of these tests require the use of a helper script or manual > > > > intervention to restart the server. > > > > > > > > Scott Mayhew (4): > > > > nfs4.1: add some reboot tests > > > > nfs4.1: add some more reboot tests > > > > nfs4.1: still more reboot tests > > > > nfs4.1: test delayed reclaim following a server reboot > > > > > > > > nfs4.1/server41tests/st_reboot.py | 284 +++++++++++++++++++++++++++++- > > > > 1 file changed, 278 insertions(+), 6 deletions(-) > > > > > > > > -- > > > > 2.17.2
(Note: These patches go on top of "nfs4.1: clean up the session and client created in Environment.init()" from Jeff Layton sent on March 14th). These patches add the following reboot tests: REBT2x: Server reboot with N clients reclaiming opens. REBT3x: Server reboot with N clients reclaming opens. Server reboots again after half the clients have reclaimed. REBT4x: Server reboot with N clients reclaiming opens. Half the clients attempt to reclaim twice. (where x in {a, b, c}, with 'a' using 10 clients, 'b' using 100 clients, and 'c' testing 1000 clients) REBT5: Server reboot where the client starts reclaiming shortly before the end of grace. Note all of these tests require the use of a helper script or manual intervention to restart the server. Scott Mayhew (4): nfs4.1: add some reboot tests nfs4.1: add some more reboot tests nfs4.1: still more reboot tests nfs4.1: test delayed reclaim following a server reboot nfs4.1/server41tests/st_reboot.py | 284 +++++++++++++++++++++++++++++- 1 file changed, 278 insertions(+), 6 deletions(-)