I'm wondering if the LMDB database has been tested on HP-UX with the Itanium processor? We are trying to use the database there and are seeing strange errors that don't occur on other platforms. Examples are assertions when trying to do cross process access and values not making it into the database correctly.
I can provide more details, but I thought I would ask about the current status first.
Kristian Amlie wrote:
I'm wondering if the LMDB database has been tested on HP-UX with the Itanium processor?
Yes, but not in the past couple years.
We are trying to use the database there and are seeing strange errors that don't occur on other platforms. Examples are assertions when trying to do cross process access and values not making it into the database correctly.
Did you define the correct CACHELINE size?
I can provide more details, but I thought I would ask about the current status first.
On 19/02/15 16:32, Howard Chu wrote:
Kristian Amlie wrote:
I'm wondering if the LMDB database has been tested on HP-UX with the Itanium processor?
Yes, but not in the past couple years.
We are trying to use the database there and are seeing strange errors that don't occur on other platforms. Examples are assertions when trying to do cross process access and values not making it into the database correctly.
Did you define the correct CACHELINE size?
I hadn't, but I tried it now. Seems to have no effect, but I suppose CACHELINE size would only affect performance? It's anyway a good tip, I will enable it for HP-UX regardless.
Any idea what else it could be?
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
Kristian Amlie wrote:
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
On 20/03/15 20:48, Howard Chu wrote:
Kristian Amlie wrote:
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
On 23. mars 2015 09:02, Kristian Amlie wrote:
On 20/03/15 20:48, Howard Chu wrote:
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
Anything else we can do here?
Have you had a chance to test it out? If you don't have access to that HP-UX version, we might be able to get a temporary login for you, Howard.
Kristian Amlie wrote:
On 23. mars 2015 09:02, Kristian Amlie wrote:
On 20/03/15 20:48, Howard Chu wrote:
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
Anything else we can do here?
Have you had a chance to test it out? If you don't have access to that HP-UX version, we might be able to get a temporary login for you, Howard.
I don't have 11.2x anywhere any more, so that would be helpful, yes. Aside from that, these are pretty old releases to still be supporting aren't they?
Howard Chu hyc@symas.com schrieb am 08.04.2015 um 12:02 in Nachricht
[...]
I don't have 11.2x anywhere any more, so that would be helpful, yes. Aside from that, these are pretty old releases to still be supporting aren't they?
We are no longer using HP-UX, but I think there was hardware where the 11i v3 (11.31) release would not run, so you had to stay at the older release...
Ulrich
Kristian Amlie wrote:
On 20/03/15 20:48, Howard Chu wrote:
Kristian Amlie wrote:
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
According to the HPUX release notes, the Unified File Cache is a new feature in HP-UX 11i v3, so 11.23 simply isn't going to work.
www.filibeto.org/unix/hp-ux/lib/rel/11.31/relnotes-5991-6469.pdf
Page 63, page 231.
Howard Chu wrote:
Kristian Amlie wrote:
On 20/03/15 20:48, Howard Chu wrote:
Kristian Amlie wrote:
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
According to the HPUX release notes, the Unified File Cache is a new feature in HP-UX 11i v3, so 11.23 simply isn't going to work.
www.filibeto.org/unix/hp-ux/lib/rel/11.31/relnotes-5991-6469.pdf
Page 63, page 231.
(Actually, it will work in WRITEMAP mode.)
On 09/04/15 03:17, Howard Chu wrote:
Howard Chu wrote:
Kristian Amlie wrote:
On 20/03/15 20:48, Howard Chu wrote:
Kristian Amlie wrote:
This took some time, but I've finally been able to make a reproducible test case. The attached test case fails on HP-UX on Itanium every time, but passes on Linux. As far as I can tell, it is a minimal test case, I cannot remove anything and still have it fail, yet I don't see any wrong API usage in it. The database should be removed before each run.
How is it best to proceed? Should I submit ticket in the bug tracker?
I have compiled and run your test case on
bash-4.2$ uname -a HP-UX malunga B.11.31 U ia64 0905345786 unlimited-user license
and there is no error. What OS revision are you using? Please provide full system description.
bash-4.3$ uname -a HP-UX sovmh356 B.11.23 U ia64 1001733034 unlimited-user license
According to the HPUX release notes, the Unified File Cache is a new feature in HP-UX 11i v3, so 11.23 simply isn't going to work.
www.filibeto.org/unix/hp-ux/lib/rel/11.31/relnotes-5991-6469.pdf
Page 63, page 231.
(Actually, it will work in WRITEMAP mode.)
Thanks Howard, that's very helpful! We will switch to WRITEMAP mode on HP-UX then.
openldap-technical@openldap.org
-
Howard Chu -
Kristian Amlie -
Ulrich Windl