[dev.icinga.com #3855] ORA-22835 causes blank status data cronks #1050
Comments
Updated by mfriedrich on 2013-04-06 22:19:31 +00:00
is this fixable for 1.9? the proposed core/ido code lives now in git branch mfriedrich/clob - https://git.icinga.org/?p=icinga-core.git;a=shortlog;h=refs/heads/mfriedrich/clob if not, #3412 can't be merged for core 1.9 release. |
Updated by Tommi on 2013-04-07 09:37:08 +00:00 the problem comes from direct fetching lob contents (this is was OCI_RETURN_LOBS does)instead of fetching the descriptor afterwards via OCI-Lob calls(read or load) using the descriptor.
It should be mentioned we have already long_output column as clob atleast for a year.
**** read clob value in smaller chunks
**** replace field value with $lobvalue |
Updated by Tommi on 2013-04-07 19:22:23 +00:00
after some hours of digging into this funny icingaOracle.php stuff i modified it by adding a new fetchlob function and dropped the existing to_char query modification. It lives on top of the current next branch in tdressler/issue_3855. Now the error is gone and i can see the service status data even with a very long output. Pls test it too. Unfortunally, no change applied to hoststatus and status_map cronk. There are "internal errors" popping up, but i couldn't found any error messages regarding this into the logs or syslog, even after applying debug setings suggested in https://wiki.icinga.org/display/testing/Icinga+Web+Testing#IcingaWebTesting-DatabaseConnection. Here i suspect another issue not related to this one. Only one problem logged
which looks more like a cleanup issue and shouldn't stop only the data esp. for this particular type. I would like to handover this problem to the web developers. Its not an Oracle issue (or someone point me to this) . If needed, i can open a new issue. |
Updated by mfriedrich on 2013-04-08 19:12:55 +00:00
i'll try to test it this week. |
Updated by mfriedrich on 2013-04-12 19:11:28 +00:00
for some reason i cannot resolve right now, i am unable to connect to my local oracle XE install, while all other tests work fine.
works when called via webbrowser.
output on the dev console with the json 500
|
Updated by Tommi on 2013-04-12 20:31:24 +00:00 your doctrine dsn doesn't work. try simailar this. note the count of slashes for each section.
|
Updated by mfriedrich on 2013-04-13 10:42:39 +00:00
buha. thanks a lot. i already read about the SID for XE being //localhost/XE and not just XE in case of such resolution errors, but passing that in that way onto the dsn is ugly. on the icingaOracle.php it's already proven that db_host and db_port are fully ignored and just placeholders here (like ido2db does). so having the correct sid seems pretty damn hard for test cases. i'll add it to my personal documentation then, thanks again. |
Updated by mfriedrich on 2013-04-13 10:53:37 +00:00 ok, reproducing the error now.
|
Updated by mfriedrich on 2013-04-13 11:01:01 +00:00 now trying your fixed branch. Throws the same errors on e.g. /icinga-web/modules/cronks/viewproc/icinga-service-template/json
|
Updated by mfriedrich on 2013-04-13 11:12:28 +00:00
I've got no idea in this field (php and oci fetching) so i'll reassign it to you. and since this might be a showstopper - if this cannot be solved til next week tuesday evening, we'll reschedule it for 1.10 |
Updated by Tommi on 2013-04-13 13:15:57 +00:00
This error i doesn't had before, but dont tried all screens. This is not related to php/oci but pure sql error. Cant believe this works well in a current version. you can't do a distinct over a dataset containing a clob field. This distinct query doesnt make sense anyway because of the nature of the sum of all selected fields there are no duplicates expected. Remove the distinct and the ora-930 is gone. BTW: perfdata is already a clob sine 1.7., means this problem is still in place since a year. i am still looking for the log which provides deeper information about "internal Errors". It is neither in "../log/icinga-web**.log", nor in ".../log/debug**.log" nor in /var/log/messages. It is very annoying to get error popups but not being able to find any details about the reason |
Updated by mfriedrich on 2013-04-13 13:23:06 +00:00 i've only used your git branch for testing the web, so you may reproduce the error easily with your install. for testing icinga web and the asynchronous json request 500, try the wiki - https://wiki.icinga.org/display/testing/Icinga+Web+Testing#IcingaWebTesting-DebugRequestsandResponses |
Updated by Tommi on 2013-04-13 19:08:15 +00:00
OK, could find the source of the query causing the ORA-00932 and removed distinct in the xml.
|
Updated by mfriedrich on 2013-04-13 19:50:41 +00:00 the problem with your distinct fix is that this was done for another fix (at least for the services recently).
there are more distinct selects around in this area. |
Updated by mfriedrich on 2013-04-13 19:57:06 +00:00 the statusmap bug was fixed in recent next commits, where your branch isn't rebased against right now. |
Updated by mfriedrich on 2013-04-13 20:11:42 +00:00 regarding the clobs and distinct - since this adds a larger historical tree, i've opened #3965 as release critical bug. i'll wait for marius to discuss what to do further. |
Updated by mfriedrich on 2013-04-16 09:12:07 +00:00 resolving #3965 will allow a proper merge then. already rebased the branch against next, will merge after the core rebase-merge is done. thanks for fixing. |
Updated by mfriedrich on 2013-04-16 09:25:17 +00:00
merged to 'next'. |
This issue has been migrated from Redmine: https://dev.icinga.com/issues/3855
Created by Tommi on 2013-03-13 21:04:11 +00:00
Assignee: Tommi
Status: Resolved (closed on 2013-04-16 09:25:17 +00:00)
Target Version: 1.9
Last Update: 2013-04-16 09:25:17 +00:00 (in Redmine)
with #3412 we introduced clob instead of varchar for serveral long columns as suggested by gunnar in #3325. With the current git master i got a lot of internal errors regarding oci fetch errors causing status screens and others to be blank
This means you have to bind a clob data type or truncate data. see http://www.php.net/manual/de/function.oci-bind-by-name.php
You can check this with the sample check from #3324. similar i found already in #2584
Changesets
2013-04-07 18:34:42 +00:00 by Tommi 68f0fdc3c175158a5447788b10ec50c920855117
2013-04-13 18:55:39 +00:00 by Tommi 9360d4a1276b4a694a096187fcb915ed3b0a27d9
2013-04-15 13:35:48 +00:00 by Tommi bd883d0
Relations:
The text was updated successfully, but these errors were encountered: