PubCrawler's WWW-Service: Errors and Problems

This page lists errors and problems that occured in connection with PubCrawler's WWW-Service.

24-26 May, 2003
- Problem: no access, limited number of jobs carried out
- Reason: network outage
- Solution: wait for restoration of network service
- Description:
  A network failure prevented access to and from the PubCrawler server over the weekend. Only a limited number of queries were carried out for Saturday, Sunday and Monday.
  For quick updates on your queries, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
21-22 January, 2003
- Problem: no access, limited number of jobs carried out
- Reason: server crash
- Solution: manual restart
- Description:
  The server was hanging for a night because of NFS problems. On January 22nd only those jobs particularly scheduled for that day of the week (Wednesday) were carried out.
  For quick updates on your queries, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
24-26 August, 2002
- Problem: no access, no jobs carried out
- Reason: server crash
- Solution: manual restart
- Description:
  Due to a crash the web server went down early in the morning of August 24th. Unfortunately it couldn't be switched on again until August 26th because of the weekend.
  For quick updates on your queries, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
02-04 January, 2002
- Problem: no access, no jobs carried out
- Reason: power failure
- Solution: manual restart
- Description:
  Due to a powerfailure the web server shut down early in the morning of January 2nd. Unfortunately it didn't come back up automatically. Because of the holidays it was only switched back on in the afternoon of January 4th.
  For quick updates on your queries, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
16/17 December, 2001
- Problem: no access, no jobs carried out
- Reason: college network down
- Solution: wait for service / reboot
- Description:
  Due to maintenance work on the college network the service got interrupted on Sunday and caused a server freeze, which was only recovered on Monday. Most of the scheduled jobs for both days were not carried out.
  For quick updates on your queries, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
13 December, 2000, 1.00 - 10.00 am GMT
- Problem: server congested
- Reason: network failure
- Solution: reboot
- Description:
  Due to network problems during the night the main server overloaded and couldn't start the queries scheduled for today. The service is back to normal now.
  If you would like updates on your queries today, please use Quickstart to run your PubCrawler jobs manually.
  
  Apologies for any inconvenience caused!
13/14 August, 2000
- Problem: server down
- Reason: harddrive failure
- Solution: replacement
- Description:
  A complete harddrive failure lead to a system crash on Sunday. The drive was replaced and most of the affected date restored.
  In case you notice that some of your results are missing or corrupted, please use Quickstart to run your PubCrawler jobs manually. Also don't forget to check the links for hits from previous days.
  
  For any questions or problems please write to pubcrawlerREMOVECAPShelp@gmail.com.
13 August, 2000, 3.30 - 12.30 pm IST
- Problem: server down
- Reason: not yet identified
- Solution: -
- Description:
  The PubCrawler web server stalled early on Sunday morning and wasn't rebooted until midday. Some of the scheduled queries will therefore be executed delayed.
  Until the cause of the problem has been identified there might be further interruptions.
  
  We are very sorry for any inconvenience caused by this!
11 August, 2000, 2.30 - 3.30 pm IST
- Problem: server down
- Reason: disk write failure
- Solution: file system check and repair
- Description:
  During the execution of the scheduled queries some disk I/O errors were detected. This required disconnection from the web and a thorough check of the affected hard drive.
  After several errors were found and repaired the system was put into operation again. If the problem should occur again the disk might need to be replaced to prevent further disruptions and/or loss of data.
  
  Sorry for the service interruption!
04/05 July, 2000
- Problem: no web access; no runs, no hits, no e-mails from PubCrawler
- Reason: default router down
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  The default router for the network segment in college on which PubCrawler is operating crashed on Tuesday evening and wasn't restarted until the next morning. During that time the PubCrawler WWW-Service wasn't accessible and didn't carry out any queries.
  People whose results weren't updated can start their jobs manually by using Quickstart.
  
  Please accept our sincerest apologies for any inconvenience caused.
18 - 20 March, 2000
- Problem: PubCrawler finds all hits from the past
- Reason: NCBI's Entrez doesn't recognize relentrezdate parameter anymore
- Solution: We are waiting for it.
- Description:
  PubCrawler searches are normally restricted to a certain relative date which is specified by the variable relentrezdate. Since the last few days this doesn't seem to be recognized by NCBI's Entrez anymore. Therefore PubCrawler retrieves all hits for each query and presents them as new hits.
  We have informed NCBI about this misbehaviour and hope that it will be fixed soon.
  
  Apologies for any inconvenience caused.
8/9 March, 2000
- Problem: no web access; no runs, no hits, no e-mails from PubCrawler
- Reason: server crashed
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  The server crashed and needed to be restarted, which took a bit longer than usual because the main administrator was not around.
  People whose results weren't updated can start their jobs manually by using Quickstart.
  Apologies for any inconvenience caused.
1 March, 2000
- Problem: no web access; no runs, no hits, no e-mails from PubCrawler
- Reason: problem with college network
- Solution: When network connection back up, use Quickstart to start your PubCrawler jobs manually.
- Description:
  The college network went down for a few hours in the morning. No PubCrawler jobs could be carried out during that time.
  People whose results weren't updated can start their jobs manually by using Quickstart.
  Apologies for any inconvenience caused.
23 February, 2000
- Problem: no web access; no runs, no hits, no e-mails from PubCrawler
- Reason: server halt
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  The server halted early in the morning which resulted in some jobs not being run.
  People whose results weren't updated can start their jobs manually by using Quickstart.
  Sorry for any inconvenience caused.
10-14 February, 2000
- Problem: no web access; no runs, no hits, no e-mails from PubCrawler
- Reason: technical problems while swapping PubCrawler jobs to other machines
- Solution: Wait for web access and use Quickstart to start your PubCrawler jobs manually.
- Description:
  Due to the huge expansion of PubCrawler within the last weeks we have to swap some of the user jobs to other machines. This process is causing some technical problems which leads to occasional interruption of the service.
  People whose results weren't updated can start their jobs manually by using Quickstart.
  Please bear with us and accept our sincerest apologies for any inconvenience caused.
27,28 January, 2000
- Problem: no runs, no hits, no e-mails from PubCrawler
- Reason: data corruption, script failure
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  Data corruption occured (probably due to hardware failure) and caused unforseen problems for the script that automates the PubCrawler jobs. None of the jobs scheduled for Thursday and Friday were carried out. They can be started manually by using Quickstart.
  Our sincerest apologies for any inconvenience caused.
26 January, 2000
- Problem: no runs, no hits, no e-mails from PubCrawler
- Reason: Server froze
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  The PubCrawler server froze this morning. Reason is not known, yet. Might be a hardware failure. Worked fine after reboot. Will investigate into this as soon as possible. Some of the runs scheduled for today (Wednesday) weren't carried out. They can be started manually by using Quickstart.
25 January, 2000
- Problem: no runs, no hits, no e-mails from PubCrawler
- Reason: power cut early in the morning
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  Due to a power cut PubCrawler shut down early in the morning. Most of the runs scheduled for today (Tuesday) weren't carried out. They can be started manually by using Quickstart.
26 - 27 October, 1999
- Problem: no direct network connection to www.ncbi.nlm.nih
- Reason: problem with name server
- Solution: Use of proxy / use Quickstart to start your PubCrawler jobs manually.
- Description:
  After the network was up again (see below) it wasn't possible to connect directly to NCBI's web-site from our server. Several users who tried to quickstart their PubCrawler jobs got failure messages in return. We temporarily routed queries through Trinity's proxy server until the problem got fixed..
24 - 26 October, 1999
- Problem: no results for some users
- Reason: network down from Sunday, 24th October, 5 pm till Tuesday, 26th October, 12:30 pm
- Solution: Use Quickstart to start your PubCrawler jobs manually.
- Description:
  The segment of the Trinity College network on which we are located was down from Sunday afternoon on. Since Monday was a holiday in Ireland the problem wasn't fixed until Tuesday lunch. Within this time no connection to NCBI could be established. Users whose jobs were executed at that time got an error message in their results file. They are encouraged to start their jobs individually through the Quickstart page or wait for the next scheduled run.
  These network failures are very annoying and completely beyond our control. We apologize for any inconveniences caused by this!
22nd October, 1999
- Problem: delayed/no results for some users
- Reason: network down from 6 - 10 a.m.
- Solution: quickstart/continued remaining runs after network was working again
- Description:
  The TCD network was down from 6 - 10 a.m. and no connection to NCBI could be established. Users whose jobs were executed at that time got an error message in their results file. They are encouraged to start their jobs individually through the Quickstart page or wait for the next scheduled run.
18 - 20th September, 1999
- Problem: no runs, no hits, no e-mails from PubCrawler
- Reason: power cut and network problems over the weekend
- Description:
  Due to a power cut PubCrawler was down from Friday 19:30 GMT to Saturday 16:37 GMT. No PubCrawler runs have been carried out on Saturday! Subsequent network problems caused many runs to result in zero hits with messages Possible network failure on Sunday and Monday.
06th September, 1999
- Problem: delayed results for some users
- Reason: power cut early in the morning
- Solution: continued remaining runs manually in the morning
- Description:
  Due to a power cut in the morning (7:45) the server was down for a while and execution of PubCrawler jobs was interrupted. After reboot the PubCrawler runs were continued.
04th August, 1999
- Problem: no results for some users / double e-mails
- Reason: network failure during the night
- Solution: restarted affected runs in the morning
- Description:
  The network was down during the night and users whose PubCrawler jobs were executed during that time would get no results and the message Possible network failure in their output file.
  The jobs of these users have been restarted after the network was working again. Those who normally get results or notification by mail will receive a second e-mail for that day.
30th July, 1999
- Problem: execution of runs delayed
- Reason: power cut over night, server didn't reboot automatically
- Solution: rebooted manually in the morning
until June 5, 1999
- Problem: for some searches not all relevant documents were found
- Reason: bug in PubCrawler / misunderstanding of query handling at PubMed
- Solution: editing of the PubCrawler script: removal of [all fields]
- Description:
  PubCrawler queries like horizontal gene transfer [All Fields] don't produce any results at PubMed although articles containing these words in their title exist. The problem seems to occur when no MeSH-term for the query words exists. When the field specifier ([All Fields]) is left out the query results in a lot of hits. Therefore from 5th of June on the specifier [All Fields] is removed from a query to retrieve all relevant documents. This might result in a particularly high number of hits for the first time your PubCrawler job is carried out after these changes. We are working at a solution to handle these cases correctly for other field specifiers, too. Apologies for any inconveniences caused!
May 25, 1999
Still problems with links. Decided to process results from NCBI and prepend server base (http://www.ncbi.nlm.nih.gov/) to each link (starting with /htbin-post). Thus we can leave out the BASE HREF tag and all page-internal references will work fine.
May 23 - 24, 1999
- Problem: NCBI-links for full reports are not working properly
- Reason: Change in PubCrawler script
- Solution: Further editing of the script and manual editing of results pages on Monday, May 24th 1999
- Description:
  The links for new hits in results pages generated at these dates are not working properly. The reason is that we modified the PubCrawler script to exclude all header information from the original NCBI pages included in the results page (there should be only one header per HTML-page and some browsers/mail-clients got confused). With this we accidentally removed a 'BASE HREF' tag necessary to retrieve documents at NCBI from within the results page. This has been corrected on Monday, 24th and was also manually added to results generated earlier that day. Nevertheless, results that were received by e-mail contained the bad links.
  As a result of the changes index links (for each results section) and back-to-top links need an absolute address now and will unfortunately not work properly for results that are sent via e-mail. Instead of pointing to the section or anchor on the same (local) page they are pointing to the according section in the original output page (on our server). At the moment there is no solution to that problem and we would welcome any suggestions on how to solve it.

For questions or remarks, please mail PubCrawler.

[PubCrawler Home Page] [PubCrawler WWW-Service] [PubCrawler Configurator] [PubCrawler Program]

Last modified at $Date: 2003/05/26 21:35:18 $
by Karsten Hokamp